Tag: AI Tools

  • Claude vs GPT vs Gemini: Coding Benchmark Leaderboard (June 2026)

    Claude vs GPT vs Gemini: Coding Benchmark Leaderboard (June 2026)

    Last verified: June 13, 2026

    As of June 13, 2026, the four models most often compared for coding work are Claude Fable 5 and Claude Opus 4.8 from Anthropic, GPT-5.5 from OpenAI, and Gemini 3.1 Pro from Google. This page is a leaderboard built on one rule: every score below is taken from a vendor’s own page or the benchmark’s official model card that we fetched on the verification date, or it is marked as not published. Several vendors publish their benchmark tables as images rather than machine-readable text; where we could not read an official figure directly, we list the metric as not machine-verifiable and link to the source document instead of estimating. The result is a smaller table than most roundups, but every number in it is one you can click through and check.

    Models and pricing (verified specs)

    These columns are confirmed from each vendor’s official model documentation. Claude prices, context windows, and cutoffs come from Anthropic’s models overview and the AWS Bedrock model card; GPT-5.5 from OpenAI’s developer docs; Gemini 3.1 Pro from Google’s DeepMind model card and the Gemini API pricing page.

    Model API ID Input / Output (per Mtok) Context Max output Knowledge cutoff
    Claude Fable 5 claude-fable-5 $10 / $50 1M 128K Not stated on overview*
    Claude Opus 4.8 claude-opus-4-8 $5 / $25 1M 128K Jan 2026
    GPT-5.5 gpt-5.5 $5 / $30 1,050,000 128K Dec 1, 2025
    Gemini 3.1 Pro gemini-3.1-pro-preview $2 / $12 (≤200K)** 1M 64K Not stated on model card

    *Anthropic’s models overview lists Fable 5’s specs and price but does not publish a knowledge-cutoff date for it in the table we fetched. **Gemini 3.1 Pro uses tiered pricing: $2 / $12 per Mtok for prompts up to 200K tokens, rising to $4 / $18 for prompts above 200K tokens (Google AI pricing page). GPT-5.5 pricing rises to 2x input / 1.5x output above 272K input tokens (OpenAI developer docs). Claude Opus 4.8 offers an optional fast mode at $10 / $50 per Mtok (Anthropic).

    Coding benchmark scores (primary-source only)

    Each cell is either a figure we read directly from a primary source on June 13, 2026, or marked “not machine-verifiable” with the source you should consult. A blank-equivalent entry never means zero — it means the official figure was not available in readable form during verification. Note the harness and version differences called out in the footnotes: they make cross-vendor cells not strictly comparable.

    Benchmark Claude Fable 5 Claude Opus 4.8 GPT-5.5 Gemini 3.1 Pro
    SWE-bench Verified Not machine-verifiable (see system card) Not machine-verifiable (see system card) Not published in retrievable primary source 80.6%
    SWE-bench Pro (Public) Not machine-verifiable (see system card) Not machine-verifiable (see system card) Not published in retrievable primary source 54.2%
    Terminal-Bench Not machine-verifiable (see system card) Not machine-verifiable (see system card) 83.4% (v2.1, Codex CLI harness)† 68.5% (v2.0, Terminus-2 harness)
    LiveCodeBench Pro Not published in retrievable primary source Not published in retrievable primary source Not published in retrievable primary source 2887 Elo

    †GPT-5.5’s Terminal-Bench 2.1 figure of 83.4% is the score Anthropic attributes to GPT-5.5 “with the Codex CLI harness” in a footnote on its Claude Opus 4.8 announcement page. It is a competitor-reported comparison, not a number we read from OpenAI directly. Google reports Gemini 3.1 Pro on Terminal-Bench 2.0 under the Terminus-2 harness (68.5%); because the version and harness differ, the Gemini and GPT-5.5 Terminal-Bench cells are not directly comparable. Gemini’s SWE-bench Verified (80.6%), SWE-bench Pro Public (54.2%), and LiveCodeBench Pro (2887 Elo) are single-attempt figures from Google’s official Gemini 3.1 Pro model card.

    What we could not verify from a primary source

    Anthropic publishes its coding comparison tables for Claude Opus 4.8 and Claude Fable 5 as images inside its announcement pages, and the full Claude Opus 4.8 System Card PDF exceeded our fetch size limit, so we could not machine-read those percentages on the verification date. OpenAI’s GPT-5.5 announcement page returned an access error to our fetcher, and its developer-docs model page lists specs and pricing but no benchmark scores. We have therefore left Claude’s and GPT-5.5’s SWE-bench figures out of the table rather than reproduce numbers we could not confirm at the source. For those figures, consult the primary documents linked in our source list: the Claude Opus 4.8 System Card, the Claude Fable 5 and Mythos 5 announcement, and OpenAI’s GPT-5.5 page. If you are choosing a model today, the verified spec table above (price, context, output, cutoff) is the part you can rely on without caveat.

    How to read a coding leaderboard

    Three cautions apply to any 2026 coding comparison. First, harness matters: the same model scores differently on Terminal-Bench depending on whether it runs under Terminus-2, a Codex CLI scaffold, or a vendor’s internal agent, which is why we annotate every Terminal-Bench cell. Second, version matters: “Terminal-Bench 2.0” and “Terminal-Bench 2.1” are different test sets, and “SWE-bench Pro” public and full splits differ — a single percentage with no version is close to meaningless. Third, a headline score is one slice of behavior; long-horizon agentic coding, tool-call reliability, and context handling over a long session often decide real-world usefulness more than a single pass rate. Treat the verified cells here as a starting point, then test the shortlist on your own repository.

    Which model has the highest published coding benchmark score in June 2026?

    We cannot crown a single winner from primary sources alone, because Anthropic and OpenAI publish their coding scores in formats we could not machine-verify on June 13, 2026. From figures we could read directly, Google’s Gemini 3.1 Pro model card reports 80.6% on SWE-bench Verified and 54.2% on SWE-bench Pro (Public). Anthropic’s and OpenAI’s comparable figures are in their system cards and announcement pages, which we link in the sources; we did not reproduce them here because they were not readable at the source during verification.

    What does Claude Fable 5 cost, and how is it different from Opus 4.8?

    Claude Fable 5 (claude-fable-5) is priced at $10 per million input tokens and $50 per million output tokens, with a 1M-token context window and up to 128K output tokens (Anthropic models overview). Claude Opus 4.8 (claude-opus-4-8) is the Opus-tier flagship at $5 / $25 per Mtok, also 1M context and 128K output, with a January 2026 knowledge cutoff. Fable 5 is Anthropic’s most capable widely released model; Opus 4.8 is the lower-priced model most teams will use for everyday agentic coding.

    Why are some benchmark cells marked “not machine-verifiable” instead of showing a number?

    Because this page only prints scores we could confirm from a primary source on the verification date. Several vendors render their benchmark tables as images, and one large system-card PDF exceeded our fetch limit, so the underlying percentages were not readable to us. Rather than copy figures from third-party trackers, we mark the cell and point you to the official document. It keeps the leaderboard honest at the cost of being shorter.

    How do the context windows compare?

    Claude Fable 5, Claude Opus 4.8, and Gemini 3.1 Pro each offer a 1M-token context window; GPT-5.5 offers 1,050,000 tokens. Maximum output is 128K tokens for Claude Fable 5, Claude Opus 4.8, and GPT-5.5, and 64K tokens for Gemini 3.1 Pro. Note that Claude Opus 4.8’s context window is 200K on Microsoft Foundry specifically, per Anthropic’s documentation.

    Is Terminal-Bench comparable across these models?

    Not cell-for-cell. Google reports Gemini 3.1 Pro on Terminal-Bench 2.0 under the Terminus-2 harness (68.5%), while the GPT-5.5 figure we show (83.4%) is Terminal-Bench 2.1 under a Codex CLI harness, as attributed by Anthropic. Different versions and different harnesses mean the two numbers should not be read as a head-to-head result.


  • Claude Skills vs MCP vs Connectors vs Plugins: What Each One Is (2026)

    Claude Skills vs MCP vs Connectors vs Plugins: What Each One Is (2026)

    Last verified: June 13, 2026

    The simplest way to keep these straight: Skills teach Claude how to do a task, MCP servers and Connectors give Claude access to external systems, Plugins bundle several of these together, and Hooks and slash commands control a Claude Code session. A Skill is a folder of instructions Claude reads when relevant. MCP (the Model Context Protocol) is an open standard that connects Claude to your tools and data. A Connector is Anthropic’s packaging of a remote MCP server inside the Claude apps. A Plugin packages any combination of commands, agents, MCP servers, hooks, and skills for Claude Code. Every definition below is taken verbatim from Anthropic’s official documentation, fetched on the verification date.

    The one-glance comparison

    This is the liftable summary. Each row is one mechanism; the third column is the distinction people most often get wrong — whether the thing teaches Claude how to do something or gives Claude access to something.

    Type What it is Teaches-HOW or gives-ACCESS Where it runs How you install / enable it
    Skill A modular capability that packages instructions, metadata, and optional resources (scripts, templates) in a SKILL.md file that Claude uses automatically when relevant. Teaches HOW. Provides domain-specific expertise: workflows, context, and best practices (procedural knowledge). In Claude’s code execution environment / VM, where Claude has filesystem access, bash, and code execution. claude.ai: upload a zip under Settings > Features. API: upload via the Skills API (/v1/skills) with the required beta headers (code-execution-2025-08-25, skills-2025-10-02, files-api-2025-04-14). Claude Code: a SKILL.md directory under ~/.claude/skills/ or .claude/skills/.
    MCP server An implementation of the Model Context Protocol — “an open-source standard for connecting AI applications to external systems.” Described as “a USB-C port for AI applications.” Gives ACCESS. Connects Claude to data sources, tools, and workflows so it can access information and perform tasks. Local (stdio) servers run as processes on your machine; remote servers run over HTTP (recommended) or SSE (deprecated). In Claude Code: claude mcp add, at local, project, or user scope. Also configurable in .mcp.json or imported from Claude Desktop / claude.ai.
    Connector A feature that “let[s] Claude access your apps and services, retrieve your data, and take actions within connected services.” Custom connectors use remote MCP. Gives ACCESS. Same access role as MCP — a Connector is the in-app packaging of a remote MCP server. Custom connectors are reached from Anthropic’s cloud infrastructure, not from your local machine. In the Claude apps under Customize > Connectors (or the in-chat “+” menu). Add a directory connector, or “Add custom connector” by URL.
    Plugin “A lightweight way to package and share any combination of” Claude Code customizations. Both — it’s a container. Bundles things that teach HOW (commands, skills) and things that give ACCESS (MCP servers), plus hooks and subagents. In Claude Code — “they’ll work across your terminal and VS Code.” The /plugin command (public beta). For a marketplace: /plugin marketplace add user-or-org/repo-name, then install from the /plugin menu.
    Hook “User-defined shell commands, HTTP endpoints, or LLM prompts that execute automatically at specific points in Claude Code’s lifecycle.” Controls behavior. Provides deterministic control rather than relying on the LLM to decide. In Claude Code, firing at lifecycle events (e.g. PreToolUse, PostToolUse, UserPromptSubmit, SessionStart, Stop). Configured in JSON settings files such as ~/.claude/settings.json or .claude/settings.json, or bundled in a plugin.
    Slash command A command starting with / that controls a Claude Code session. Includes built-ins (e.g. /help, /compact) and custom commands. Teaches HOW (custom) / controls session (built-in). Custom commands have been merged into Skills. In the Claude Code session (terminal or VS Code). Built-ins ship with Claude Code. Custom: a Markdown file under .claude/commands/ (project) or ~/.claude/commands/ (personal); a .claude/skills/<name>/SKILL.md does the same.

    Skills: teaching Claude a procedure

    An Agent Skill is “a directory containing a SKILL.md file” with YAML frontmatter plus instructions, and optionally additional markdown files, executable scripts, and reference resources. The point is procedural knowledge — it turns “general-purpose agents into specialists” by giving Claude the workflows and best practices for a task, the way you’d write an onboarding guide for a new teammate. Anthropic ships pre-built Skills for PowerPoint, Excel, Word, and PDF, and you can author your own.

    What makes Skills cheap to install in bulk is progressive disclosure: Claude loads information in stages instead of all at once. The numbers below come straight from Anthropic’s Skills overview.

    Loading level When loaded Token cost (per Anthropic docs) Content
    Level 1: Metadata Always, at startup ~100 tokens per Skill name and description from the YAML frontmatter
    Level 2: Instructions When the Skill is triggered Under 5k tokens The SKILL.md body — workflows and guidance
    Level 3+: Resources As needed Effectively unlimited Bundled files read or executed via bash without loading their contents into context

    The name field is capped at 64 characters (lowercase letters, numbers, hyphens; it cannot contain the reserved words “anthropic” or “claude”), and the description is capped at 1,024 characters. One important constraint: custom Skills do not sync across surfaces — a Skill uploaded to claude.ai is not automatically available via the API, and Claude Code Skills are filesystem-based and separate from both.

    MCP: the open standard for access

    The Model Context Protocol is, in Anthropic’s words, “an open-source standard for connecting AI applications to external systems.” The canonical analogy: “Think of MCP like a USB-C port for AI applications. Just as USB-C provides a standardized way to connect electronic devices, MCP provides a standardized way to connect AI applications to external systems.” Using MCP, “AI applications like Claude or ChatGPT can connect to data sources (e.g. local files, databases), tools (e.g. search engines, calculators) and workflows (e.g. specialized prompts).”

    An MCP server can expose three kinds of building block — tools, resources, and prompts. In Claude Code, resources are referenced with @server:protocol://resource/path and prompts surface as commands in the form /mcp__servername__promptname. You connect a server with claude mcp add, choosing a transport and a scope:

    Scope Loads in Shared with team Stored in
    Local (default) Current project only No ~/.claude.json
    Project Current project only Yes, via version control .mcp.json in project root
    User All your projects No ~/.claude.json

    For transports, HTTP is “the recommended option for connecting to remote MCP servers,” local stdio servers “run as local processes on your machine,” and SSE is explicitly marked deprecated in favor of HTTP.

    Connectors: MCP, packaged for the apps

    A Connector is how the Claude apps surface MCP. Per Anthropic’s help center, “Connectors let Claude access your apps and services, retrieve your data, and take actions within connected services,” and “Custom connectors using remote MCP are available on Claude, Cowork, and Claude Desktop.” So a Connector is not a different technology from MCP — a custom connector is a remote MCP server wired into the Claude UI.

    The most consequential detail is where the connection originates: “Custom connectors (remote MCP servers) are reached from Anthropic’s cloud infrastructure, not from your local machine.” That means a custom-connector MCP server must be reachable over the public internet — one hosted only on a private network, behind a VPN, or blocked by a firewall will not connect even if you can reach it yourself.

    Aspect Directory (pre-built) connector Custom connector
    Source Pre-built integrations in the Connectors Directory Added by you via a remote MCP server URL
    Plan availability Available across Claude plans Free, Pro, Max, Team, and Enterprise
    Free-plan limit Per directory “Free users are limited to one custom connector.”
    Where to add it Customize > Connectors, or the in-chat “+” > Connectors > Manage connectors

    Plugins: a bundle, not a single thing

    A Plugin is “a lightweight way to package and share any combination of” Claude Code customizations. The official announcement lists four bundle components, and the Claude Code documentation adds skills as a fifth thing a plugin can carry:

    Component What it adds
    Slash commands Custom shortcuts for frequently-used operations
    Subagents Purpose-built agents for specialized development tasks
    MCP servers Connections to tools and data sources through MCP
    Hooks Customizations of Claude Code’s behavior at key workflow points
    Skills Per the Claude Code docs, a plugin can include a skills/ directory; plugin skills use a plugin-name:skill-name namespace

    You install a plugin “directly within Claude Code using the /plugin command.” To pull from a marketplace — “curated collections where other developers can discover and install plugins” — you run /plugin marketplace add user-or-org/repo-name and then install from the /plugin menu. Plugins “work across your terminal and VS Code.” Plugins were announced on October 9, 2025, as a public beta for all Claude Code users.

    Hooks and slash commands: controlling the session

    The last two mechanisms aren’t about adding capability — they’re about controlling a Claude Code session. Hooks are “user-defined shell commands, HTTP endpoints, or LLM prompts that execute automatically at specific points in Claude Code’s lifecycle.” Their defining property is determinism: they provide deterministic control rather than relying on the LLM to make decisions. A PreToolUse hook can, for example, block a destructive rm -rf command regardless of what Claude intended. Hooks are configured in JSON settings files (such as ~/.claude/settings.json or a project’s .claude/settings.json) and fire at events including SessionStart, UserPromptSubmit, PreToolUse, PostToolUse, and Stop.

    Slash commands start with / and control the session. Built-in commands like /help and /compact ship with Claude Code. Custom commands are Markdown files — a project command lives at .claude/commands/<name>.md and a personal one at ~/.claude/commands/<name>.md, with the file name becoming the command. As of the 2026 Claude Code docs, custom commands have been merged into Skills: a file at .claude/commands/deploy.md and a skill at .claude/skills/deploy/SKILL.md “both create /deploy and work the same way,” and existing .claude/commands/ files keep working.

    Is a Connector the same as an MCP server?

    Effectively yes, for custom connectors. Anthropic states “Custom connectors using remote MCP are available on Claude, Cowork, and Claude Desktop,” and that they “are reached from Anthropic’s cloud infrastructure.” A Connector is the Claude-app packaging of a remote MCP server; MCP is the underlying open standard.

    What’s the difference between a Skill and an MCP server?

    A Skill teaches Claude how to do a task — it “provide[s] Claude with domain-specific expertise: workflows, context, and best practices.” An MCP server gives Claude access to external systems — it connects Claude “to data sources, tools and workflows.” One is procedural knowledge; the other is a connection.

    Do Skills cost a lot of context tokens?

    Not until used. Per Anthropic’s docs, Level 1 metadata costs about 100 tokens per Skill and is always loaded; the full SKILL.md body (under 5k tokens) only loads when the Skill is triggered; and bundled resources are read on demand with effectively no upfront cost. This is the “progressive disclosure” design.

    What can a Claude Code plugin contain?

    “Any combination of” slash commands, subagents, MCP servers, and hooks, per the announcement; the Claude Code documentation adds that a plugin can also bundle a skills/ directory. You install one with the /plugin command, optionally from a marketplace added via /plugin marketplace add.

    Are custom slash commands still a thing?

    They still work, but they’ve been folded into Skills. The Claude Code docs state custom commands “have been merged into skills,” that existing .claude/commands/ files keep working, and that a command file and an equivalent SKILL.md both produce the same / command. Skills add optional extras like supporting files and automatic invocation.


  • Migrating Off Retired Claude Models: The Breaking-Change Checklist (2026)

    Migrating Off Retired Claude Models: The Breaking-Change Checklist (2026)

    Last verified: June 13, 2026

    Claude Opus 4 (claude-opus-4-20250514) and Claude Sonnet 4 (claude-sonnet-4-20250514) are deprecated and retire on June 15, 2026, after which requests to them return a 404. The official replacements are claude-opus-4-8 and claude-sonnet-4-6. But swapping the model string alone will break a working integration: depending on which target you choose, several request parameters that were valid on the May 2025 models now return a 400 error, and two changes alter behavior silently. This page maps each removed or changed parameter to the exact failure and the fix.

    One distinction governs the whole migration. The Opus path (to claude-opus-4-8) is the strict one: it removes temperature/top_p/top_k and manual thinking budgets entirely. The Sonnet path (to claude-sonnet-4-6) is gentler: it keeps sampling parameters (with the older “one of temperature or top_p, not both” rule) and still accepts budget_tokens as deprecated-but-functional. The one rule both paths share: assistant-turn prefills now return 400.

    The breaking-change matrix

    Each row is a change that breaks on at least one migration target. “Error” means the API rejects the request server-side (HTTP 400) even though the SDK request type still type-checks. “Silent” means no error — the behavior simply differs.

    Change On Opus 4.8 On Sonnet 4.6 Symptom Fix
    thinking: {type:"enabled", budget_tokens:N} 400 error (removed) Deprecated, still works 400 on Opus; cost/latency drift on Sonnet thinking: {type:"adaptive"} + output_config.effort
    temperature / top_p / top_k 400 error (removed) Keep only one of temperature or top_p 400 on Opus if any set; 400 on Sonnet if both set Remove on Opus; steer via prompt. Keep one on Sonnet
    Assistant-turn prefill (last message role:"assistant") 400 error 400 error Request rejected on both output_config.format (structured outputs) or system-prompt instruction
    thinking.display default Defaults to "omitted" Returns summarized text Reasoning text empty on Opus (silent) Set display: "summarized" on Opus
    Tokenizer New tokenizer (more tokens) Unchanged tokenizer Same text counts higher on Opus; max_tokens too tight Re-baseline with count_tokens; add headroom
    output_format (top-level) Deprecated API-wide Deprecated API-wide Works, but slated for removal Move to output_config: {format: {...}}

    Model ID swaps and retirement dates

    Retiring model Model ID Retires Replacement
    Claude Opus 4 claude-opus-4-20250514 (alias claude-opus-4-0) June 15, 2026 claude-opus-4-8
    Claude Sonnet 4 claude-sonnet-4-20250514 (alias claude-sonnet-4-0) June 15, 2026 claude-sonnet-4-6

    These are the original May 2025 models, not the later Opus 4.6 or Sonnet 4.5 releases. Use the exact replacement strings above — do not append a date suffix to claude-opus-4-8 or claude-sonnet-4-6 (they are dateless pinned snapshots).

    budget_tokens to adaptive thinking

    The Opus path removes the fixed thinking budget. thinking: {type:"enabled", budget_tokens:N} returns a 400 on claude-opus-4-8. The replacement is adaptive thinking — the model decides how much to think per request — with overall depth controlled by the effort parameter (low | medium | high | xhigh | max). There is no direct token-count equivalent; effort is an output-level control, not a thinking budget.

    # Before (Claude Opus 4 / Sonnet 4)
client.messages.create(
    model="claude-opus-4-20250514",
    max_tokens=16000,
    thinking={"type": "enabled", "budget_tokens": 10000},
    messages=[{"role": "user", "content": "..."}],
)

# After (Claude Opus 4.8)
client.messages.create(
    model="claude-opus-4-8",
    max_tokens=16000,
    thinking={"type": "adaptive"},
    output_config={"effort": "high"},  # or "max", "xhigh", "medium", "low"
    messages=[{"role": "user", "content": "..."}],
)

    On the Sonnet path, budget_tokens is deprecated but still functional on claude-sonnet-4-6, so it will not 400 — but you should still migrate to adaptive thinking. Note also that Sonnet 4.6 defaults to effort: "high" where Sonnet 4 had no effort parameter at all; if you do not set it explicitly you may see higher latency and token use after the swap.

    Sampling parameters: removed vs. restricted

    This is where the two paths diverge most. On claude-opus-4-8, setting temperature, top_p, or top_k to any non-default value returns a 400. Remove them entirely and steer behavior through prompting instead. (If you used temperature=0 for determinism, note it never guaranteed identical outputs on prior models either.)

    # Opus path — sampling params 400 on claude-opus-4-8
# Before
client.messages.create(
    model="claude-opus-4-20250514",
    temperature=0.7,
    top_p=0.9,
    messages=[...],
)

# After — remove them
client.messages.create(
    model="claude-opus-4-8",
    messages=[...],
)

    On claude-sonnet-4-6 the older Claude 4.x rule still applies: you may pass one of temperature or top_p, but passing both returns a 400. So a Sonnet 4 to Sonnet 4.6 move only requires dropping one of the two if you were setting both.

    Assistant-turn prefills to structured outputs

    Prefilling the final assistant turn — ending your messages array with a role: "assistant" message to force a response shape — returns a 400 on both claude-opus-4-8 and claude-sonnet-4-6. This is the one breaking change you cannot dodge by choosing the gentler target. The replacement depends on what the prefill was doing.

    Prefill was used for Replacement
    Forcing JSON / YAML / schema output output_config.format with a json_schema
    Forcing a classification label A tool with an enum field, or structured outputs
    Skipping preambles (“Here is…”) System-prompt instruction: respond directly, no preamble
    Continuing an interrupted response Move continuation into the user turn
    Steering around bad refusals Usually unnecessary now — plain user-turn prompting suffices 
    # Before (fails on both targets) — prefill forcing JSON shape
messages=[
    {"role": "user", "content": "Extract the name."},
    {"role": "assistant", "content": "{\"name\": \""},
]

# After — structured outputs replace the prefill
client.messages.create(
    model="claude-opus-4-8",
    max_tokens=1024,
    output_config={"format": {"type": "json_schema", "schema": SCHEMA}},
    messages=[{"role": "user", "content": "Extract the name."}],
)

    Thinking display: the silent one

    On claude-opus-4-8, thinking blocks still stream, but their thinking text field is empty unless you opt in — the default is display: "omitted". There is no error; if your UI rendered the summarized reasoning, it now shows a long pause before output. Restore it by setting the display mode:

    thinking = {
    "type": "adaptive",
    "display": "summarized",  # default is "omitted" on Opus 4.8/4.7
}

    The block-field name is unchanged — it is still block.thinking on a thinking-type block. The fix is the request parameter, not the response-handling code. (Sonnet 4.6 is not affected by this default change.)

    The new tokenizer: re-baseline max_tokens

    This change is Opus-only and easy to miss because it produces no error. claude-opus-4-8 uses the tokenizer introduced with Opus 4.7, under which the same text tokenizes to roughly 1x–1.35x as many tokens — up to about 35% more, around 30% on typical content, varying by workload. Three consequences:

    What to check Why
    max_tokens ceilings and compaction triggers The same output now consumes more tokens; tight limits truncate mid-thought
    Client-side token estimators (e.g. fixed char-to-token ratios) Calibrated against the old tokenizer; now undercount
    Cost and rate-limit dashboards count_tokens returns higher numbers; re-baseline before reacting

    Re-run client.messages.count_tokens(model="claude-opus-4-8", ...) on a representative sample of your prompts. Do not apply a blanket multiplier. Sonnet 4.6 keeps the older tokenizer, so a Sonnet 4 to Sonnet 4.6 move has no tokenizer re-baseline to do.

    The full checklist

    Step Opus 4 to 4.8 Sonnet 4 to 4.6
    Update model ID string Required Required
    Replace budget_tokens with adaptive thinking Required (400) Recommended (deprecated)
    Sampling params Remove all (400) Keep only one (both 400)
    Remove assistant-turn prefills Required (400) Required (400)
    Set display: "summarized" if showing reasoning Required for visible thinking Not applicable
    Re-baseline max_tokens for new tokenizer Required Not applicable
    Set effort explicitly Defaults to high Defaults to high
    Move output_format to output_config.format Recommended Recommended
    Verify tool inputs parsed with a JSON parser Recommended Recommended
    Spot-check one request, then roll out Required Required

    If you run Claude Code, /claude-api migrate applies the model swap, breaking-parameter changes, prefill replacement, and effort calibration across a codebase, then produces a verify-it-yourself checklist. It asks you to confirm scope before editing any files.

    Is migrating off Claude Opus 4 really not just a model-string change?

    No. Moving to claude-opus-4-8 also requires removing temperature/top_p/top_k and any budget_tokens (all now return 400), removing assistant-turn prefills (400), opting back into summarized thinking if your UI shows it, and re-baselining max_tokens for the new tokenizer. Only the Sonnet 4 to Sonnet 4.6 move is close to a drop-in — and even that requires removing prefills.

    When exactly do Claude Opus 4 and Sonnet 4 stop working?

    June 15, 2026. After that date, requests to claude-opus-4-20250514 and claude-sonnet-4-20250514 return a 404. These are the original May 2025 models, not Opus 4.6 or Sonnet 4.5.

    What replaces budget_tokens now that it errors on Opus?

    Adaptive thinking (thinking: {type:"adaptive"}) plus the effort parameter inside output_config. There is no exact token-count equivalent: the model decides how much to think per request, and effort (low through max) tunes overall depth and spend. On Sonnet 4.6, budget_tokens still works but is deprecated.

    Why does the same prompt cost more tokens on Opus 4.8?

    Opus 4.8 uses the tokenizer introduced with Opus 4.7, under which the same text produces roughly 1x–1.35x as many tokens (about 30% more on typical content, up to ~35%). Re-run the count_tokens endpoint against claude-opus-4-8 and give max_tokens and compaction triggers extra headroom. Sonnet 4.6 keeps the older tokenizer, so it is unaffected.

    My thinking summaries disappeared after migrating to Opus — is that a bug?

    No. On Opus 4.8 (and 4.7), thinking.display defaults to "omitted", so thinking blocks stream with an empty text field. Set display: "summarized" in your thinking config to restore visible reasoning. The field name is unchanged; only the default flipped.


  • I Actually Used Claude Fable 5 Before the Government Pulled It. Here’s What They Took.

    I Actually Used Claude Fable 5 Before the Government Pulled It. Here’s What They Took.

    Three days. That’s how long Claude Fable 5 existed in the wild before the US government killed it.

    On Monday, June 9, Anthropic launched Fable 5 and Mythos 5. On Thursday, June 12, Commerce Secretary Howard Lutnick issued an export control directive ordering Anthropic to suspend access for any foreign national. Since Anthropic can’t verify nationality in real time, they shut it down for everyone. Globally. Immediately. The stated reason was a narrow jailbreak vulnerability — one Anthropic says exists in other publicly deployed models too.

    I’m not writing this to debate export controls. I’m writing this because I spent those three days running Fable 5 in production — not benchmarking it, not kicking the tires, actually building with it — and I have something most people writing about this don’t have: receipts.

    Day One: The Model Dropped and I Put It to Work

    Fable 5 launched June 9. By that afternoon, I had it running a Batch 8 sprint across my Tygart Media site — refreshing 10 pages of Claude content that needed updating. Fable 5 updated comparison tables, corrected model names across the lineup, added FAQPage schema, injected internal links, and expanded word counts. Post 4787 went from 750 words to 1,602. Post 9821 went from 1,782 to 2,543. Five posts refreshed with full SEO treatment — schema, FAQs, RankMath meta, silo links — in a single session.

    That same day, I had Fable 5 write a complete guide to itself. Not a press release rewrite — a 2,100-word article with an interactive cost calculator, a model picker tool, and a section called “How We Actually Use Each Model” that mapped my real production workflows to each tier: Haiku for the daily 25-post SEO sweeps, Sonnet for desk articles, Opus for deep refreshes, Fable for portfolio-wide audits and strategy. The draft landed in Notion with scoped CSS and JS, ready to paste into WordPress as a single Custom HTML block.

    Day Two: Fable 5 Ran My Entire SEO Audit

    June 10. I ran a full SEO audit of tygartmedia.com through Fable 5. It identified that Fable 5 itself was the top content gap — a model launched 24 hours ago with zero dedicated coverage and peak search intent. So it wrote the article to fill its own gap. It drafted the piece, tagged the slug, assigned the category, and queued internal links to five existing posts.

    That same day, Fable 5 wrote and published “The Signal: AI Just Split Into Two Lanes” — a 1,400-word field notes piece that wove together Fable 5’s launch, OpenAI’s S-1, Chrome WebMCP, and the emerging thesis that AI was splitting into a product lane and an infrastructure lane. The article went through the full pipeline: SEO optimization, AEO with 8 FAQ Q&As, GEO entity enrichment, Article + FAQPage schema, taxonomy assignment, internal linking, quality gate — then published via REST API. It even created the LinkedIn draft in Metricool and scheduled it for 2:30 PM Pacific.

    That article exists right now at tygartmedia.com. I didn’t write it. Fable 5 did, with me directing the strategy and approving the output. The quality bar was real journalism, not AI slop.

    Day Three: Building the Infrastructure Layer

    June 11. While the Fable 5 Complete Guide sat in Notion waiting for a featured image, I was using Fable 5 to build the systems that would keep my content operation running. I had it update the Claude Intelligence Desk — my Notion page that serves as the authoritative source of truth for every Claude model name, API string, and price across my entire content operation. Every article gets verified against that desk before publishing. Fable 5 updated it with its own pricing: $10 input, $50 output per million tokens.

    I also had Fable 5 design my Pricing Freshness Engine — a WordPress mu-plugin that shadow-checks Anthropic’s live pricing against what’s displayed on my site. The engine had been running in shadow mode since June 2, catching drift before it reaches readers. Fable 5 added itself to the canonical pricing store.

    Meanwhile, my 6 scheduled email agent tasks — morning triage, midday check, afternoon wrap, newsletter extraction, weekly prep, and weekly self-audit — were running on the same Claude infrastructure, handling my inbox while I focused on building. The whole system runs on my Max plan. No extra API charges.

    What Fable 5 Actually Felt Like

    Here’s what the benchmarks don’t tell you: Fable 5 understood intent, not just instructions.

    When I told it to run a page refresh, it didn’t just update the text — it checked model names against my Intelligence Desk, verified pricing against live documentation, added schema markup, expanded FAQs, injected internal links, and updated the dateline. It treated each task as a system, not a checklist.

    When I asked it to write the Complete Guide, it included a section about how we actually use each model tier in production — because it knew from context that an article about Claude models on a site that runs on Claude models should demonstrate firsthand expertise, not just recite specs. It even built interactive JavaScript widgets inline — a cost calculator and a model picker — without being asked, because it understood the article needed to be useful, not just informative.

    The gap between Fable 5 and what came before it was the largest single-model jump I’ve experienced since I started building on Claude in 2024.

    What Most Commentators Are Missing

    Most people writing about the shutdown never used Fable 5. They’re debating precedent, policy, the implications for AI regulation. All valid. But the conversation is incomplete without understanding what was actually deployed.

    This is the first time the US government has aimed export controls at a deployed commercial AI model rather than at chips or hardware. That’s unprecedented. Anthropic complied but publicly disagreed, calling it a likely misunderstanding based on a narrow jailbreak that exists in other models too.

    Every other Claude model — Opus, Sonnet, Haiku — remains fully available and unaffected.

    What I Lost

    Here’s what the government took from me specifically:

    My Fable 5 Complete Guide is sitting in Notion, ready to publish, with the proxy fix queued. The pricing pages need Fable 5 rows added. The Freshness Engine needs Fable 5 in its canonical store. The WordPress proxy’s ALLOWED_DOMAINS needs a one-line gcloud update. All of it was queued up. All of it was dependent on a model that no longer exists.

    The infrastructure I built this week — the Intelligence Desk, the Pricing Freshness Engine, the content pipeline that ran “The Signal” from draft to published with schema and social scheduling in a single session — all of that still works with Opus and Sonnet. But the ceiling is lower. The tasks that Fable 5 handled in one pass will take two or three with the models that remain.

    What Happens Now

    Anthropic says this isn’t permanent. They’re working to restore access.

    For people like me who build businesses on top of these tools, the uncertainty is the real cost. Three days is long enough to build production workflows, deploy infrastructure, and write articles that reference a model’s existence — and short enough that all of it gets yanked before you can publish.

    But I’m not pulling back. This week confirmed the trajectory. AI at this level isn’t a nice-to-have — it’s the infrastructure of how modern knowledge work gets done. Whether it’s Fable 5 or whatever comes after it, this capability exists now. You can’t un-ring that bell.

    I know because I rang it. For three days, I built real things with a model the government decided the world shouldn’t have. And the work is still there in my Notion, waiting.

    Will Tygart is the founder of Tygart Media, where he builds AI-native content operations across a portfolio of WordPress sites. He has been building production workflows on Claude since 2024. His Claude Intelligence Desk, Pricing Freshness Engine, and content pipeline systems were all built or upgraded using Claude Fable 5 during its three-day window.

  • Latest Claude Models — June 2026 (Current Lineup, Pricing, and Specs)

    Latest Claude Models — June 2026 (Current Lineup, Pricing, and Specs)

    Updated June 12, 2026. Fable 5 is the current top-tier model, released June 9, 2026. The full lineup: Fable 5 → Opus 4.8 → Sonnet 4.6 → Haiku 4.5. Pricing and availability verified against Anthropic’s official docs.

    Current Claude Models — Quick Reference

    Model API ID Input $/MTok Output $/MTok Context Best For
    Claude Fable 5 🆕 claude-fable-5 $10.00 $50.00 1M tokens Complex engineering, long-horizon agentic work
    Claude Opus 4.8 claude-opus-4-8 $5.00 $25.00 1M tokens Everyday advanced work, high-volume pipelines
    Claude Sonnet 4.6 claude-sonnet-4-6 $3.00 $15.00 1M tokens Balanced capability and speed
    Claude Haiku 4.5 claude-haiku-4-5-20251001 $1.00 $5.00 200K tokens High-speed, cost-sensitive tasks

    All four models support vision, tool use/function calling, and batch processing. Fable 5 and Opus 4.8 are available on AWS Bedrock, Google Cloud Vertex AI, and Microsoft Azure AI Foundry in addition to the direct Anthropic API.

    Claude Fable 5 (June 9, 2026)

    Fable 5 is Anthropic’s first publicly available Mythos-class model — a capability tier previously restricted to research and select enterprise partners. It’s the most capable model Anthropic has released to date.

    What makes Fable 5 different:

    • SWE-bench Verified: 95.0% (vs 88.6% for Opus 4.8)
    • SWE-bench Pro: 80.0% (vs 69.2%)
    • Senior Engineer benchmark: 91/100 (vs ~63/100)
    • Adaptive extended thinking (always on, not a mode switch)

    Important limitations:

    • 2x the cost of Opus 4.8 ($10/$50 vs $5/$25)
    • Mandatory 30-day data retention — not available under zero data retention (ZDR)
    • Safety classifiers route cybersecurity, biology, chemistry, and distillation prompts to an Opus 4.8 fallback — you pay Fable 5 rates for Opus 4.8 output in those domains
    • Higher latency on complex tasks (60 seconds to several minutes vs 3–15 seconds for Opus 4.8)

    Free through June 22, 2026: Claude Pro, Max 5x, Max 20x, Team, and Enterprise subscription plans include Fable 5 at no extra charge during the launch window.

    Claude Opus 4.8

    Opus 4.8 is Anthropic’s current workhorse for serious work — the right default for most API applications and Claude Code use. It supports zero data retention (ZDR), which Fable 5 does not.

    Key specs:

    • Context: 1M tokens
    • Max output: 32K tokens per request
    • Extended thinking: Available (opt-in mode)
    • ZDR: Yes
    • Batch API: Yes (50% discount on batch processing)

    Use Opus 4.8 as your default model unless you have a specific reason to go up to Fable 5 or down to Sonnet/Haiku. It hits the best balance of capability, speed, cost, and data policy flexibility.

    Claude Sonnet 4.6

    Sonnet 4.6 targets use cases where response speed matters and you don’t need Opus-level reasoning. It’s the model Anthropic’s own infrastructure runs Claude.ai on for Pro subscribers doing day-to-day chat work.

    Key specs:

    • Context: 1M tokens
    • Max output: 64K tokens per request
    • Extended thinking: Available
    • ZDR: Yes
    • Typical latency: 1–5 seconds for most tasks

    Good for: content generation pipelines, customer-facing chat, document analysis at volume, anything where sub-5-second response time matters.

    Claude Haiku 4.5

    Haiku 4.5 is Anthropic’s fastest and cheapest model. At $1 input / $5 output per million tokens, it’s 10x cheaper than Fable 5.

    Key specs:

    • Context: 200K tokens (smaller than the Opus/Sonnet/Fable 1M window)
    • Max output: 16K tokens per request
    • Latency: Sub-second for most tasks
    • ZDR: Yes

    The 200K context window is the main limitation. For tasks that fit within that window — classification, short-form generation, routing, extraction — Haiku 4.5 is the cost-optimal choice. For longer documents or conversations, step up to Sonnet or Opus.

    How to Choose the Right Claude Model

    The decision framework I use:

    1. Does the task require multi-step reasoning, complex coding, or long-horizon autonomy? → Fable 5 (if cost and latency are acceptable) or Opus 4.8
    2. Is this a routine task at reasonable volume? → Opus 4.8 as the default
    3. Does latency matter more than maximum reasoning depth? → Sonnet 4.6
    4. Is this high-volume, short-context, cost-sensitive work? → Haiku 4.5
    5. Does your use case require zero data retention? → Any model except Fable 5

    Most production applications use a routing strategy: Fable 5 or Opus 4.8 for the hard jobs, Haiku 4.5 for classification and pre-processing, Sonnet 4.6 for user-facing response generation.

    Claude Subscription Plans and Model Access (June 2026)

    Plan Price Models Included
    Free $0 Limited Sonnet 4.6 access
    Pro $20/mo ($17 annual) Sonnet 4.6, Opus 4.8, Fable 5 (through June 22)
    Max 5x $100/mo All models, 5x usage vs Pro
    Max 20x $200/mo All models, 20x usage vs Pro
    Team Standard $20/seat/mo (annual), $25 month-to-month All models + admin features
    Team Premium $100/seat/mo (annual), $125 month-to-month All models + priority + advanced admin
    Enterprise $20/seat + usage at API rates (contact sales) All models + ZDR + custom retention + SSO

    Claude Code (the CLI tool) is included in all paid subscription plans. API access for building your own applications is separate — billed per token via the Anthropic Console regardless of subscription status.

    Legacy Models (Still Available, No Longer Latest)

    These models are still available via the API but are not Anthropic’s current recommended versions:

    • Claude Opus 4.7 (claude-opus-4-7) — prior Opus tier, succeeded by 4.8
    • Claude Opus 4.6 (claude-opus-4-6) — two generations back
    • Claude Sonnet 4.5 (claude-sonnet-4-5) — prior Sonnet tier
    • Claude 3.5 Haiku / Sonnet / Opus — Claude 3.x generation, still functional for legacy integrations

    If you’re building a new application, start with the current lineup. Legacy model IDs are useful for maintaining compatibility in existing applications that haven’t been updated.

    Platform Availability

    Platform Fable 5 Opus 4.8 Sonnet 4.6 Haiku 4.5
    Anthropic API (direct) ✓ ✓ ✓ ✓
    AWS Bedrock ✓ ✓ ✓ ✓
    Google Cloud Vertex AI ✓ ✓ ✓ ✓
    Microsoft Azure AI Foundry ✓ ✓ ✓ ✓
    GitHub Copilot ✓ (via Foundry) ✓ ✓ —

    Frequently Asked Questions

    What is the newest Claude model?
    As of June 2026, Claude Fable 5 is the newest and most capable model Anthropic has released. It launched June 9, 2026. The API model ID is claude-fable-5.

    Is Claude Fable 5 the same as Claude 5?
    No. Anthropic changed the naming convention — there is no “Claude 5.” The Fable series is a new tier above the Opus/Sonnet/Haiku hierarchy. Fable 5 is Anthropic’s first Mythos-class model released for general availability.

    What is the most powerful Claude model?
    Claude Fable 5 is currently the most powerful. For tasks where Fable 5’s safety classifier routing applies (cybersecurity, biology, chemistry, distillation), or where zero data retention is required, Claude Opus 4.8 is the appropriate top-tier choice.

    What Claude model does Claude.ai use by default?
    Depends on your plan. Free tier uses a limited version of Sonnet. Pro and Max subscribers access Opus 4.8 as the default with Fable 5 available (through June 22, 2026 included, after that plan-dependent). Claude.ai routes to the appropriate model for your plan automatically.

    How do I use the latest Claude model in the API?
    Set the model parameter in your API request to the model ID. For Fable 5: "model": "claude-fable-5". For Opus 4.8: "model": "claude-opus-4-8". See the full API reference at console.anthropic.com.

    What’s the difference between Claude Opus 4.8 and Fable 5?
    Fable 5 is significantly stronger on complex engineering tasks — SWE-bench Pro: 80% vs 69.2%, and Senior Engineer benchmark: 91 vs ~63 out of 100. The trade-off: Fable 5 costs 2x more ($10/$50 vs $5/$25 per MTok), has higher latency, and requires 30-day data retention. For most work, Opus 4.8 is the right choice. Full Fable 5 vs Opus 4.8 breakdown here.

    Changelog

    • June 12, 2026 — Added Claude Fable 5 (released June 9). Updated pricing table. Added platform availability table.
    • May 2026 — Claude Opus 4.8 and Sonnet 4.6 were the current top-tier models.

    This page is updated as Anthropic releases new models. Last verified: June 12, 2026. For API pricing, check console.anthropic.com — the canonical source.

  • Claude Code Getting Started: Installation, First Run, and the 5 Commands You’ll Use Daily

    Claude Code Getting Started: Installation, First Run, and the 5 Commands You’ll Use Daily

    Claude Code is Anthropic’s official CLI for Claude — a terminal-based agent you can point at any codebase and have it read, write, test, and ship code. It’s different from the Claude.ai chat interface in one key way: Claude Code can act, not just answer. It reads your actual files, runs your actual commands, and makes changes that stick.

    This guide walks you through installation, first run, and the commands that cover 90% of what you’ll do daily.

    What Claude Code Is (and Isn’t)

    Claude Code runs in your terminal. It gives Claude access to your local machine — file system, shell, and any MCP servers you configure — so it can do real engineering work: implement features, fix bugs, write tests, explain unfamiliar codebases, and run multi-step agentic workflows.

    It is not a code autocomplete plugin (that’s what GitHub Copilot does). Claude Code is a conversational agent that works at the task level, not the token level. You describe what you want; it figures out the steps and executes them.

    Installation

    Claude Code requires Node.js 18 or later. Install via npm:

    npm install -g @anthropic-ai/claude-code

    Verify the install:

    claude --version

    That’s the only dependency. Claude Code is a Node.js CLI — no Docker, no Python env, no platform-specific setup beyond Node.

    First-Run Authentication

    The first time you run claude, it walks you through authentication. You have two options:

    Option 1: Claude subscription (Pro, Max, Team, Enterprise)
    Run claude, select “Login with Claude.ai,” and it opens a browser window to authorize. Your subscription covers Claude Code usage — no separate API billing.

    Option 2: Anthropic API key
    Set your API key as an environment variable before running:

    export ANTHROPIC_API_KEY="sk-ant-..."
claude

    Or on Windows:

    $env:ANTHROPIC_API_KEY = "sk-ant-..."
claude

    API key usage is billed per token at standard API rates. For heavy daily use, a Max subscription ($100–$200/month) is usually more economical than API billing.

    Your First Session

    Navigate to a project directory and start Claude Code:

    cd ~/projects/my-app
claude

    Claude Code reads your directory automatically. At the > prompt, describe what you want:

    > What does this codebase do? Give me a 3-paragraph overview.

    Claude reads the files it needs and responds. No configuration required for basic usage — Claude Code infers context from the directory you’re in.

    The 5 Commands You’ll Use Daily

    1. claude — Start an interactive session

    claude

    Launches the REPL (read-eval-print loop). This is where you spend most of your time. Claude has access to your current directory’s files, can run bash commands, and can call any MCP servers you’ve configured.

    Within a session, you can:

    • Ask questions about the codebase
    • Request implementations (“add a rate limiter to the auth middleware”)
    • Have Claude run tests and fix failures
    • Use /help to see available slash commands
    • Use /clear to reset context without leaving the session
    • Press Escape twice to interrupt a running task

    2. claude -p "prompt" — One-shot non-interactive mode

    claude -p "What are all the API endpoints in this codebase?"

    Runs a single prompt and exits. No REPL. Good for scripting, CI pipelines, or quick one-off queries you don’t want to interrupt a workflow for. Output goes to stdout — pipe it wherever you need it.

    claude -p "Summarize the changes in the last 10 commits" | pbcopy

    3. claude mcp add — Connect an external tool

    claude mcp add github -- npx -y @modelcontextprotocol/server-github

    Adds an MCP server to your Claude Code configuration. After running this, Claude can call the server’s tools in any session. Common additions:

    # File system access (scoped to a directory)
claude mcp add files -- npx -y @modelcontextprotocol/server-filesystem ~/Documents

# GitHub integration
claude mcp add github -- npx -y @modelcontextprotocol/server-github

# Web search
claude mcp add search -- npx -y @modelcontextprotocol/server-brave-search

    The GitHub and Brave Search servers need API tokens — set them as environment variables before the server starts, or pass them via the --env flag in the mcp add command.

    4. claude -c — Continue the last conversation

    claude -c

    Resumes your most recent Claude Code conversation, including all prior context. Essential for multi-session work on a feature. If you closed the terminal mid-task, claude -c picks up exactly where you left off.

    For a specific prior conversation:

    claude --resume SESSION_ID

    5. claude --model — Select the model for a session

    claude --model claude-opus-4-8

    Claude Code defaults to the most capable available model for your plan. You can override this per session. Current options:

    • claude-fable-5 — Highest capability, complex tasks (2x cost vs Opus 4.8)
    • claude-opus-4-8 — Default for most work, strong balance of quality and speed
    • claude-sonnet-4-6 — Faster responses, good for routine tasks
    • claude-haiku-4-5-20251001 — Fastest, lowest cost, short tasks

    Slash Commands Inside a Session

    While in a Claude Code session (> prompt), these slash commands are available:

    Command What It Does
    /help Show all available commands
    /clear Clear conversation context (keep the session open)
    /compact Compress prior context to save tokens while preserving essential memory
    /cost Show token usage and estimated cost for the current session
    /model Switch the model mid-session
    /review Request a multi-agent code review of the current branch
    /init Generate a CLAUDE.md file with project context for this repo
    /exit End the session

    CLAUDE.md — Project-Level Context

    Drop a CLAUDE.md file in your project root and Claude Code reads it automatically at session start. Use it to encode project-specific context Claude shouldn’t have to re-derive every session:

    # My Project

## Architecture
- Backend: FastAPI + PostgreSQL
- Frontend: React + TypeScript
- Deployed to: AWS ECS

## Development
- Tests: `pytest tests/`
- Local server: `./scripts/start-dev.sh`
- Database migrations: `alembic upgrade head`

## Rules
- Never modify migration files directly
- All API routes go in `src/routes/`
- Use `httpx` not `requests` for HTTP calls

    Generate a starter CLAUDE.md for an existing project with /init.

    Permission Modes

    Claude Code asks for confirmation before running bash commands, creating files, or making other changes — unless you grant it broader permissions. There are three ways to control this:

    • Default: Claude asks before each tool use that modifies files or runs commands
    • --dangerously-skip-permissions: Skip all confirmations. Use only in isolated environments (Docker containers, CI). Not for everyday use on your primary machine.
    • Session-level allowlist: During a session, you can approve individual tools for the rest of the session by selecting “Allow always” when prompted

    For most work, the default confirmation behavior is the right trade-off — it keeps you in the loop on changes without requiring you to pre-define a permission policy.

    IDE Integration

    Claude Code integrates with VS Code and JetBrains IDEs. Install the extension from each marketplace, then launch Claude Code from inside the IDE. This keeps the terminal panel visible alongside your editor without alt-tabbing between windows.

    The IDE extensions also add shortcuts for common actions like opening Claude Code in the current file’s directory and running one-shot queries against the selected code.

    Frequently Asked Questions

    What’s the difference between Claude Code and Claude.ai?
    Claude.ai is the web chat interface — good for questions, document analysis, and writing. Claude Code is a terminal CLI that can access your local files, run commands, and act autonomously on multi-step tasks. Claude.ai can’t modify files on your machine; Claude Code can.

    Does Claude Code cost extra on top of my Claude subscription?
    No. Claude Pro, Max, Team, and Enterprise subscriptions include Claude Code access. You use the same account. Heavy agentic usage counts toward the plan’s usage limits, but there’s no separate Claude Code fee.

    Can Claude Code access the internet?
    Not by default. Claude Code’s built-in WebFetch tool can fetch content from a specific URL when you provide it. For live web search, add the Brave Search or similar MCP server. Claude can’t browse freely without explicit tool access.

    What does Claude Code do with my code?
    Claude Code sends the file contents and context it needs to the Anthropic API for inference. Standard Anthropic API data policies apply — if you’re using an API key, you can configure zero data retention. If you’re using a subscription, default Anthropic retention policies apply. Review Anthropic’s privacy policy for current details.

    Is Claude Code open source?
    Claude Code itself (the CLI client) is not open source — it’s an Anthropic product. The MCP server ecosystem it connects to includes many open-source servers, and the MCP specification itself is open.

    What version of Node.js do I need?
    Node.js 18 or later. Run node --version to check. The Long-Term Support (LTS) version is always a safe choice.

    Last verified: June 12, 2026. Claude Code is updated frequently — run npm update -g @anthropic-ai/claude-code to stay current.

  • What Is Model Context Protocol (MCP)? The Complete Guide for Claude Users

    What Is Model Context Protocol (MCP)? The Complete Guide for Claude Users

    Model Context Protocol (MCP) is the reason Claude can read your files, query your database, search the web, and push code to GitHub — all from inside a single conversation. Without it, Claude would be limited to whatever you paste in manually. With it, Claude connects to almost any external system.

    Quick answer: MCP is an open standard developed by Anthropic that lets AI models securely connect to external tools, data sources, and services through a standard client-server architecture. You install an MCP server for the system you want Claude to access. Claude becomes a client that calls that server. The server executes the action and returns results.

    The Problem MCP Solves

    Before MCP, connecting an AI model to external data meant one of two things: either the AI company built a native integration (slow, expensive, proprietary), or you cobbled together a pipeline that passed data manually between systems.

    Neither approach scales. If Claude natively supported every database, every API, every file format, and every SaaS tool on the planet, the model would be perpetually behind. And manual copy-paste workflows aren’t agentic — they require you to do all the coordination work the AI should be doing.

    MCP solves this with a universal adapter layer. Instead of building individual integrations, Anthropic defined a standard. Now any developer can build an MCP server for any system, and any MCP-compatible AI client (like Claude) can use it automatically.

    How MCP Works

    MCP uses a client-server model over two transport mechanisms:

    • stdio: The MCP server runs as a local subprocess on your machine. Claude Code spawns it, communicates via standard input/output. This is the most common setup.
    • HTTP/SSE: The MCP server runs as a network service. Claude connects over HTTP with Server-Sent Events for streaming. Better for remote or shared servers.

    The communication protocol underneath is JSON-RPC 2.0 — a lightweight, well-understood standard for calling methods and getting results.

    Each MCP server exposes one or more of three primitives:

    • Tools: Functions Claude can call. Example: read_file(path), create_issue(title, body), run_query(sql). Claude decides when to call them based on context.
    • Resources: Data sources Claude can read. Example: the contents of a directory, a database schema, a project’s README. Resources are passive — they don’t take actions, they expose information.
    • Prompts: Reusable prompt templates that servers can provide to standardize how Claude interacts with them.

    When Claude sees a task that could benefit from an available tool, it calls the tool, receives the result, and incorporates it into the response. This happens automatically — you don’t have to tell Claude when to use MCP. Claude decides based on what the server exposes.

    MCP in Claude Code vs Claude Desktop

    Both Claude Code (the CLI tool) and Claude Desktop support MCP, but they configure servers differently.

    Claude Code

    Claude Code has built-in MCP management via the claude mcp command family:

    claude mcp add my-server -- npx -y @modelcontextprotocol/server-filesystem /path/to/directory
claude mcp list
claude mcp remove my-server

    Servers added with claude mcp add are stored in your Claude Code config (~/.claude.json or the project-level .claude/settings.json). Project-level configs let you commit MCP server setups to source control so the whole team gets them automatically.

    Claude Code also ships with a set of built-in tools that behave like MCP servers but don’t require separate installation: file read/write/edit, bash execution, glob search, grep, web fetch, and the agent spawning tools you’re reading about in this article.

    Claude Desktop

    Claude Desktop reads MCP server configuration from a JSON file:

    • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
    • Windows: %APPDATA%\Claude\claude_desktop_config.json

    A typical config entry looks like this:

    {
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "/Users/you/Documents"]
    },
    "github": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-github"],
      "env": {
        "GITHUB_PERSONAL_ACCESS_TOKEN": "ghp_your_token_here"
      }
    }
  }
}

    Restart Claude Desktop after editing the config. Each server you add appears in the Claude Desktop interface with a hammer icon, and Claude can access its tools in any conversation.

    The Most Useful MCP Servers

    Anthropic maintains a reference set of official MCP servers. These are the ones worth knowing:

    Server What It Does Package
    Filesystem Read/write files and directories on your local machine @modelcontextprotocol/server-filesystem
    GitHub Read repos, create issues, open PRs, push code @modelcontextprotocol/server-github
    PostgreSQL Read-only SQL queries against a Postgres database @modelcontextprotocol/server-postgres
    SQLite Read/write a local SQLite database file @modelcontextprotocol/server-sqlite
    Brave Search Live web search via Brave’s Search API @modelcontextprotocol/server-brave-search
    Puppeteer Headless browser — screenshot pages, scrape, fill forms @modelcontextprotocol/server-puppeteer
    Slack Read channels, send messages, search workspace @modelcontextprotocol/server-slack
    Google Drive Read and search Google Drive files @modelcontextprotocol/server-google-drive
    Git Git operations — log, diff, commit, branch management @modelcontextprotocol/server-git
    Memory Persistent key-value knowledge graph across conversations @modelcontextprotocol/server-memory

    Beyond the official set, hundreds of community-built MCP servers cover everything from Notion and Linear to AWS and Docker. The MCP ecosystem grew faster than almost anyone expected after the November 2024 launch.

    Installing Your First MCP Server

    The fastest path is Claude Code with the filesystem server. This gives Claude read/write access to a directory you specify — useful for any project work.

    Prerequisites: Node.js installed (the server runs via npx).

    In your terminal:

    claude mcp add filesystem -- npx -y @modelcontextprotocol/server-filesystem ~/Documents/projects

    That’s it. Open a Claude Code session. Claude can now list, read, write, and search files inside ~/Documents/projects. Try: “List all Python files in this directory and summarize what each one does.”

    For Claude Desktop, edit the claude_desktop_config.json file directly (see format above), then restart the app.

    What MCP Cannot Do

    A few things worth understanding before you build on MCP:

    MCP servers don’t persist between conversations. Each Claude session starts fresh. If you need state persistence, you need a server with its own storage layer (the Memory server handles this specifically).

    MCP doesn’t bypass Claude’s safety guidelines. Claude still decides whether to execute a tool call based on safety and ethics reasoning. Connecting a filesystem server doesn’t give Claude unlimited license to delete files — Claude will still confirm before destructive operations.

    Subprocess MCP servers are local. The stdio transport runs servers on your machine. This means they only work when you’re running Claude Code locally. For remote or team-shared access, you need HTTP/SSE transport with a hosted server.

    Security Considerations

    MCP servers have real permissions. The filesystem server can read and write files. The GitHub server can push code to your repos. The Postgres server can run SQL queries.

    Apply the principle of least privilege:

    • Scope filesystem servers to the directory you actually need, not /
    • Use read-only database credentials where you don’t need writes
    • Create GitHub tokens with minimum required scope (e.g., repo for private repos, not org-level admin)
    • Never commit environment variables containing API keys to source control, even in .claude/settings.json — use env var references instead

    MCP servers run with the permissions of the user running Claude. If something goes wrong with a tool call, it can have real consequences. The upside: everything runs locally and through your own credentials — there’s no MCP cloud intermediary with access to your data.

    MCP and Claude Code’s Agentic Workflows

    The full power of MCP shows up in Claude Code’s multi-step agentic mode. When Claude Code has access to git, a filesystem, a browser, and a search tool simultaneously, it can execute workflows like:

    1. Search the web for a library’s current API (Brave Search)
    2. Read your existing code to understand the integration point (filesystem)
    3. Write the updated code (filesystem write)
    4. Run tests (bash)
    5. Create a PR (GitHub)

    Each of these steps would require a separate tool in a traditional automation stack. With MCP, Claude orchestrates all of them within a single session, using whatever servers are available.

    This is what makes MCP the infrastructure layer for agentic AI — not a feature, but the foundation that makes complex AI-driven workflows possible.

    Frequently Asked Questions

    What does MCP stand for?
    Model Context Protocol. It’s an open standard for connecting AI models to external tools, data sources, and services through a standard client-server interface.

    Who created MCP?
    Anthropic created MCP and released it as an open standard in November 2024. The specification and reference servers are open-source on GitHub. While Claude is the primary client, other AI systems can implement MCP clients too.

    Do I need to install MCP to use Claude?
    No. Claude works without any MCP servers. MCP is an extension layer — you add servers when you want Claude to access specific external systems. Claude Code also ships with a set of built-in tools (file operations, bash, web fetch) that don’t require MCP installation.

    Is MCP available on Claude.ai (the web app)?
    MCP server support is primarily in Claude Desktop and Claude Code. The Claude.ai web interface has its own tool integrations (web search, document analysis) but doesn’t support custom MCP servers in the same way.

    What’s the difference between MCP tools and Claude’s native tools in Claude Code?
    Claude Code’s native tools (Read, Write, Bash, Glob, Grep, WebFetch, Agent) are built into the application and don’t require a separate server process. MCP servers are external — they run as subprocesses or network services that Claude Code connects to. Both expose tools that Claude can call; the mechanism for loading them is different.

    How do I build my own MCP server?
    Anthropic provides official SDKs for building MCP servers in TypeScript, Python, Go, and other languages. The TypeScript SDK (@modelcontextprotocol/sdk) is the most mature. Start with Anthropic’s MCP documentation and the reference server implementations on GitHub as templates.

    Last verified: June 12, 2026. MCP specification and server ecosystem evolve quickly — check the official Anthropic MCP documentation for the current spec.

  • AEO Content Optimizer — Claude AI Skill for Featured Snippets

    AEO Content Optimizer — Claude AI Skill for Featured Snippets

    Paste your article. Get back the version built to win the featured snippet.

    Who This Is For

    Built for site owners and content marketers who publish good content that never gets picked as the answer — no featured snippets, no People Also Ask placements, invisible in voice results and AI Overviews while thinner competitor pages take the box.

    The Problem

    Answer engines do not reward the best content — they reward the most extractable content. A page that buries its answer in paragraph six loses to a page that answers in the first 50 words under a question heading, formatted the way the snippet wants. Restructuring for extraction is mechanical, learnable work — and almost nobody does it. This skill does it on every piece you paste.

    What It Does

    • Performs answer-first surgery: a direct, self-contained 40–60 word answer placed immediately under each question heading
    • Converts topical headings into the question formats searchers actually use, mapped to real query variants
    • Matches the winning snippet format per query — paragraph, numbered list, or table — and rebuilds the block to fit
    • Builds a genuine FAQ section and generates the matching FAQPage JSON-LD (and warns about duplicate schema before you paste)
    • Runs a voice pass so direct answers survive a smart-speaker read
    • Returns a change log plus an honest note on what content is missing that the query demands

    What You Get

    • The aeo-content-optimizer.skill file — installs in claude.ai or Claude Code in about two minutes
    • README with installation steps and tested example prompts
    • Works on existing posts, new drafts, and competitor-gap rewrites

    $47 one-time

    Buy Now →

    Secure checkout via Square — all major cards accepted

    Want a custom version built specifically for your business? Email will@tygartmedia.com

    Frequently Asked Questions

    Do I need technical knowledge to use this?

    No. You paste your content and your target question. The skill restructures and returns paste-ready output, including the schema block.

    Does it work for my niche?

    Yes — the method is format-driven, not topic-driven. Local services, SaaS, e-commerce, professional services, and content sites all follow the same extraction rules.

    Will it change my voice or facts?

    It restructures; it does not genericize. Anything it cannot verify is flagged for you to supply rather than invented.

    How is this delivered?

    Within 24 hours of purchase via email from will@tygartmedia.com. Skill file and setup guide delivered as a ZIP download.

    Does this require a paid Claude subscription?

    Installing as a custom skill requires a paid Claude plan (Pro, $20/mo, or higher) with code execution enabled. Your download also includes a free-plan setup option — paste the skill into a Claude Project’s instructions — that works on any plan.

  • Claude Fable 5 Complete Guide

    Claude Fable 5 Complete Guide

    New in 2026

    Everything you need to know about Anthropic’s new frontier tier — pricing, context window, model comparisons, and how to route the right work to the right model.

    Updated June 2026
    ·
    ~14 min read
    ·
    Includes interactive calculators

    What Is Claude Fable 5?

    Claude Fable 5 is Anthropic’s new frontier model tier — positioned above Opus in the lineup and designed for tasks where raw capability, extended reasoning depth, and massive context handling matter more than cost. Where Opus 4.8 set the bar for complex multi-step reasoning, Fable 5 raises it with a 1-million-token context window, enhanced agentic autonomy, and improved performance on long-horizon software engineering, research synthesis, and cross-domain analysis tasks.

    The “Fable” naming signals a new generation of model architecture rather than an incremental update. Anthropic positions it as the model you reach for when a task exceeds what Opus can do reliably — not as a replacement for Opus, Sonnet, or Haiku in their respective cost tiers.

    Quick Facts — Claude Fable 5

    Context Window
    1M
    tokens (~750K words)

    Max Output
    32K
    tokens per response

    Input Price
    $10
    per million tokens

    Output Price
    $50
    per million tokens

    Cache Write
    $12.50
    per million tokens

    Cache Read
    $1.00
    per million tokens

    Key positioning: Fable 5 is the model for tasks where Opus 4.8 produces reliable but imperfect results — long codebase audits, full-document analysis, complex multi-agent orchestration, and strategic synthesis across large corpora. For most production workflows, Sonnet remains the value pick.

    Full Model Lineup Comparison

    Here’s how the complete 2026 Claude lineup stacks up across every dimension that matters for production usage:

    Model Input $/M Output $/M Context Max Out Vision Tool Use Extended Think Best For
    ◆ Fable 5 $10 $50 1M 32K ✓ ✓ ✓ Deep Max-capability tasks, 1M+ context
    ◆ Opus 4.8 $5 $25 200K 32K ✓ ✓ ✓ Complex reasoning, agentic workflows
    ◆ Sonnet 4.6 $3 $15 200K 16K ✓ ✓ ✓ Production apps, content at scale
    ◆ Haiku 4.5 $1 $5 200K 8K ✓ ✓ — High-volume, latency-sensitive tasks

    Prices are per million tokens. Cache read is 90% cheaper than standard input across all models. Batch API provides an additional 50% discount on both input and output.

    Capability Matrix — What Each Model Can Do

    Capability Fable 5 Opus 4.8 Sonnet 4.6 Haiku 4.5
    Full codebase analysis (>500K tokens) ✓ Native ⚠ Chunked ✗ ✗
    Extended thinking / chain-of-thought ✓ Deep ✓ ✓ ✗
    Multi-step agentic orchestration ✓ Best ✓ Good Limited
    Computer use ✓ ✓ ✓ ✗
    MCP tool integration ✓ ✓ ✓ ✓
    Prompt caching ✓ ✓ ✓ ✓
    Batch API (50% discount) ✓ ✓ ✓ ✓
    PDF / document analysis ✓ ✓ ✓ Limited
    Real-time streaming ✓ ✓ ✓ ✓
    Structured JSON output ✓ ✓ ✓ ✓

    Interactive Cost Calculator

    Estimate your monthly API spend across the full model lineup. Enter your token volumes below — the calculator models prompt caching and Batch API discounts automatically.

    Token Cost Calculator






    Estimated Monthly Cost
    $0.00

    Which Claude Model Should You Use?

    Answer three questions to get a model recommendation tailored to your use case.

    Model Picker — 3 Questions
    1. How large is your context? (document/codebase size)
    Under 50K tokens
    50K–200K tokens
    200K–1M tokens

    2. How complex is the task?
    Simple / structured (classify, extract, format)
    Moderate (draft, summarize, QA)
    Complex (reason, plan, code, orchestrate)

    3. How cost-sensitive is this workload?
    Very — high volume, every cent counts
    Moderate — quality matters more than cost
    Not sensitive — quality and capability first

    How We Actually Use Each Model

    These are real production workflows mapped to the right tier — built from running Claude in content operations, publishing automation, and knowledge management at scale. No hypotheticals.

    Haiku 4.5 — High Volume
    Daily SEO Refresh Pipeline
    • 25-post-per-day SEO metadata refresh
    • Article classification and tag assignment
    • Structured data extraction from web pages
    • Keyword density checks across large post archives
    • Link validation and redirect flagging
    Sonnet 4.6 — Production Default
    Editorial Content at Scale
    • Desk article writing (1,200–2,500 words)
    • Content brief execution from keyword clusters
    • FAQ and schema markup generation
    • Cross-site content adaptation and localization
    • Monthly client update drafts and summaries
    Opus 4.8 — Complex Reasoning
    Workers & Deep Refreshes
    • Agentic Notion Workers (multi-step pipelines)
    • Deep content refresh with competitive gap analysis
    • Multi-database synthesis and reporting
    • Strategy documents requiring extended reasoning
    • Code generation for automation scripts
    Fable 5 — Max Capability
    Portfolio Audits & Strategy
    • Full-site content audits (500+ posts in single context)
    • Cross-domain strategy synthesis across large corpora
    • Complex multi-agent orchestration at the flagship tier
    • Long-horizon planning requiring deep reasoning depth
    • Codebase-wide analysis and architecture review

    Routing principle: The right model is the cheapest one that reliably completes the task. Haiku handles volume. Sonnet handles production. Opus handles complexity. Fable 5 handles scale + complexity together — specifically the cases where you’d need Opus and more context than Opus can hold.

    The Economics: Routed vs All-Fable

    Smart model routing is where API costs get controlled. Here’s a real-world comparison of a mixed content-and-automation workload at scale — routed vs running everything on Fable 5.

    Workload Monthly Volume Routed Model Routed Cost All-Fable 5 Cost Savings
    SEO metadata batch refresh 750 posts/mo Haiku 4.5 + Batch $1.20 $18.75 93% less
    Article drafting 90 articles/mo Sonnet 4.6 $8.10 $67.50 88% less
    Agentic worker runs 200 runs/mo Opus 4.8 $22.50 $45.00 50% less
    Full-site portfolio audits 4 audits/mo Fable 5 $24.00 $24.00 —
    Total — Routed $55.80 $155.25 64% less

    Stacking Discounts: Caching + Batch API

    Two discount mechanisms compound independently:

    • Prompt caching: Cache your system prompt and shared context once. Subsequent requests pay ~10% of the input price for cache reads. On Fable 5, that’s $1.00/M instead of $10.00/M on cached tokens — a 90% reduction on your largest cost lever.
    • Batch API: Submit requests asynchronously (results within 24 hours) for a flat 50% discount on both input and output. Works on all four models. Best for non-real-time workloads like overnight refreshes, audits, or bulk classification.
    • Stacked: Caching + Batch combined can bring effective Fable 5 input cost from $10/M to ~$0.50/M on cached tokens — making it economically viable for high-volume tasks that previously only fit Haiku’s budget.

    See our Claude context window guide for more on how to structure prompts to maximize cache hit rates.

    Related Claude Resources

    Claude Context Window Guide
    Claude Projects Deep Dive
    Claude Code Plan Mode
    Claude MCP Setup Guide
    Opus 4.7 Feature Deep Dive
    Claude Code vs Cursor (2026)

    Claude Fable 5 FAQ

    Claude Fable 5 sits above Opus 4.8 in the lineup. The primary difference is context window size — Fable 5 offers 1 million tokens vs Opus 4.8’s 200K — and the depth of extended reasoning for highly complex tasks. Opus 4.8 remains the right choice for most complex agentic workflows at half the cost. Fable 5 is best when you need both maximum context and maximum reasoning depth simultaneously, or when a task has routinely hit the limits of what Opus can do reliably.

    Claude Fable 5 is priced at $10 per million input tokens and $50 per million output tokens — 2× Opus 4.8 ($5/$25), 3.3× Sonnet 4.6 ($3/$15), and 10× Haiku 4.5 ($1/$5). Prompt caching drops the effective input cost to $1.00/M on cache reads, and the Batch API adds a 50% discount on all tokens for non-real-time workloads. Stacking both discounts makes Fable 5 viable for higher-volume use cases than the base price suggests.

    Claude Fable 5 has a 1-million-token context window — approximately 750,000 words or roughly 1,500 pages of text. This is 5× the context window of Opus 4.8, Sonnet 4.6, and Haiku 4.5 (all 200K). In practice, a 1M context window lets you pass entire codebases, long research corpora, or full document archives in a single API call without chunking or retrieval workarounds. For more on context window mechanics, see our full context window guide.

    Yes. Claude Fable 5 is available through the Anthropic API using the model ID claude-fable-5-20260101 (check the Anthropic documentation for the exact identifier). It supports the same API surface as the rest of the Claude family — streaming, tool use, prompt caching, vision, the Batch API, and MCP server integration. Access requires an Anthropic API account with Fable 5 enabled on your usage tier.

    Fable 5 is available in Claude.ai on the Pro and Team plans. The interface lets you select it from the model picker when starting a conversation. Like Opus, Fable 5 in claude.ai has message limits that reset on a rolling window — it’s designed for individual complex tasks rather than high-volume API workloads. For production-scale usage, the API with the Batch API discount is the more economical path.

    Yes — and Fable 5’s extended thinking is the deepest in the lineup. Where Opus 4.8 supports extended thinking for complex reasoning tasks, Fable 5 uses a more capable reasoning engine designed for tasks that require longer chains of inference, more working memory, and more reliable self-correction. It’s particularly effective on math, logic, long-horizon planning, and tasks where the model needs to hold and manipulate many interdependent concepts simultaneously.

    For most content production — articles, blog posts, social copy, summaries, SEO content — Sonnet 4.6 is the right call. It produces high-quality output at 3.3× less cost than Fable 5, and for typical content lengths (500–3,000 words), the quality difference is minimal. Reach for Fable 5 when you need to synthesize across a very large corpus (e.g., auditing 200+ posts simultaneously), when the content requires deep domain reasoning that benefits from extended thinking, or when the task involves both large-context ingestion and complex output generation in a single pass.

    Three levers in order of impact: (1) Model routing — only use Fable 5 when the task genuinely requires it; route everything else to Opus, Sonnet, or Haiku based on complexity and volume. (2) Prompt caching — structure your system prompt and shared context so it can be cached; cache reads cost $1.00/M instead of $10.00/M on Fable 5. (3) Batch API — submit non-real-time workloads via the Batch API for a flat 50% discount. Stacking all three — routing + caching + batch — can reduce effective per-task costs by 85–95% compared to unoptimized Fable 5 calls.

    More Claude Guides from Tygart Media

    We run Claude in production every day. These are the guides that come from using it, not just writing about it.

    Context Window Guide
    Claude Projects
    Plan Mode
    MCP Setup
    Code vs Cursor

  • AI Model Monitoring: Tracking Identity Infrastructure

    AI Model Monitoring: Tracking Identity Infrastructure