Claude tool use (also called function calling) is the capability that transforms Claude from a conversational AI into an agentic system that can interact with external services, execute code, query databases, and take real-world actions. This guide covers how tool use works, the three execution modes, the built-in server tools, and practical implementation examples.
What Is Tool Use?
Tool use lets you define functions that Claude can call during a conversation. When Claude determines that a tool would help answer a user’s request, it generates a tool call (specifying the tool name and arguments), your code executes the function, and the result is returned to Claude to continue the conversation.
Example flow: User asks “What’s the weather in Seattle?” → Claude calls your get_weather function with {"location": "Seattle"} → Your code calls a weather API → Returns data to Claude → Claude generates a natural language response incorporating the weather data.
Defining Tools
tools = [
{
"name": "get_stock_price",
"description": "Get the current stock price for a given ticker symbol",
"input_schema": {
"type": "object",
"properties": {
"ticker": {
"type": "string",
"description": "The stock ticker symbol (e.g., AAPL, GOOGL)"
}
},
"required": ["ticker"]
}
}
]
response = client.messages.create(
model="claude-sonnet-4-6",
max_tokens=1024,
tools=tools,
messages=[{"role": "user", "content": "What's Apple's current stock price?"}]
)
The Three Execution Modes
1. Client-Side Execution
Your application receives the tool call, executes the function locally or via external APIs, and returns the result. This is the standard pattern — you control the execution environment and can call any service.
2. Server-Side Execution (Built-in Tools)
Anthropic provides built-in tools that Claude can execute server-side without your code doing anything:
web_search: Real-time web search
code_execution: Execute Python code in a sandbox
bash: Run shell commands
text_editor: Read and edit files (used in Claude Code)
3. Tool Runner SDK (Programmatic)
Anthropic’s Tool Runner SDK automates the tool call/execute/return loop, letting you build agentic workflows without writing the orchestration loop manually.
Handling Tool Results
# After receiving a tool_use block from Claude
if response.stop_reason == "tool_use":
tool_use = next(block for block in response.content if block.type == "tool_use")
tool_name = tool_use.name
tool_input = tool_use.input
# Execute your function
result = your_function(tool_input)
# Return result to Claude
follow_up = client.messages.create(
model="claude-sonnet-4-6",
max_tokens=1024,
tools=tools,
messages=[
{"role": "user", "content": "What's Apple's stock price?"},
{"role": "assistant", "content": response.content},
{"role": "user", "content": [{"type": "tool_result", "tool_use_id": tool_use.id, "content": str(result)}]}
]
)
Frequently Asked Questions
What is the difference between tool use and function calling?
They’re the same thing — Anthropic uses “tool use” as the preferred term, while “function calling” is the term OpenAI popularized. Both describe the same capability: letting an AI model invoke defined functions during a conversation.
How many tools can I define for Claude?
Claude supports up to several hundred tools in a single request, though performance is best with a focused set relevant to the task. Each tool definition consumes input tokens, so large tool sets have a cost impact.
Claude computer use is a capability that lets Claude control a computer — click buttons, type text, navigate browsers, run applications, and execute multi-step tasks as if it were a human operator. As of 2026, it’s one of the most powerful and underexplored capabilities in the Claude ecosystem. This tutorial covers what it is, how to set it up, what it’s actually useful for, and where it still falls short.
What Is Claude Computer Use?
Computer use is an API capability (not available in the standard Claude.ai interface) that lets Claude interact with a desktop environment via screenshots and tool calls. Claude sees the screen, decides what to click or type, executes that action, sees the updated screen, and continues — iterating until the task is complete.
This is different from a browser extension or web scraper. Claude is operating a real (or virtualized) computer environment the same way a human would — by looking at the screen and interacting with what it sees.
Current Benchmark Performance
On OSWorld — the leading benchmark for computer use agents — Claude currently scores around 22% task completion on the most complex tasks. ChatGPT’s computer use scores higher on this specific benchmark at approximately 75%. This gap is real and matters for production use cases requiring high reliability. For simpler, more structured tasks, Claude’s computer use performs considerably better.
Setting Up Claude Computer Use
Computer use requires API access. The basic setup:
Anthropic API key (API tier with computer use enabled)
A virtual machine or containerized desktop environment (Docker with a lightweight Linux desktop is the standard approach)
The Anthropic Python or TypeScript SDK
Anthropic provides a reference implementation with a Docker-based Ubuntu environment, a noVNC interface for monitoring, and starter code. This is the fastest path to a working computer use setup.
Best Current Use Cases
Web research and data extraction: Navigate websites, extract structured data, fill in forms — tasks that don’t have APIs
Repetitive desktop workflows: Tasks that require clicking through multiple application screens
Legacy software interaction: Applications without APIs where the only interface is visual
Key Limitations to Know
Reliability: Computer use is significantly less reliable than direct API calls for the same tasks. Where an API returns structured data, computer use can misread a screen or click the wrong element
Speed: Screenshot-based interaction is slow compared to direct integration
Cost: Each screenshot and tool call consumes API tokens; complex tasks can be expensive
Sensitive actions: Never use computer use for high-stakes irreversible actions (sending emails, making purchases) without human-in-the-loop verification
Frequently Asked Questions
Is Claude computer use available in Claude.ai?
No. Computer use is an API capability available through the Anthropic API, not the standard Claude.ai web interface.
How does Claude computer use compare to ChatGPT’s?
On OSWorld benchmarks, ChatGPT’s computer use currently leads at approximately 75% vs Claude’s ~22%. For production use cases requiring high reliability, this gap matters. Both are improving rapidly.
Jared Kaplan is the Chief Science Officer of Anthropic and one of the most consequential AI researchers alive. His 2020 paper on neural scaling laws — co-authored with Sam McCandlish and others — changed how every major AI lab thinks about model development. He is a TIME100 AI honoree, has testified before the U.S. Senate, and Forbes estimates his net worth at $3.7 billion. Yet outside of AI research circles, his name remains largely unknown to the general public.
Academic Background
Kaplan holds a PhD in physics, having trained as a theoretical physicist before pivoting to AI. Like several Anthropic co-founders, his physics background proved directly applicable to machine learning — particularly in developing the mathematical frameworks for understanding how AI systems scale. Physics training emphasizes finding simple underlying laws that explain complex phenomena, which is exactly what scaling law research does.
The Discovery That Changed AI: Scaling Laws
In January 2020, Kaplan and colleagues at OpenAI published “Scaling Laws for Neural Language Models” — a paper that demonstrated something remarkable: AI model performance improves in a smooth, predictable way as you increase model size, training data, and compute budget. The relationship follows a power law, meaning you can forecast how capable a model will be before training it, simply by knowing how much compute you’re using.
This was not merely an academic finding. It gave AI labs a roadmap: if you want a more capable model, you know roughly how much more investment is required. It directly enabled the aggressive scaling strategies that produced GPT-4, Claude 3, and every frontier model since. The paper has been cited tens of thousands of times and is considered foundational to the modern AI race.
Co-Founding Anthropic
Kaplan was among the seven OpenAI researchers who left in 2021 to found Anthropic. His technical authority — particularly in understanding what training configurations produce which capabilities — made him a natural fit as Chief Science Officer, the role he holds today.
Recognition and Public Profile
Kaplan was named to TIME’s 100 Most Influential People in AI, one of a handful of researchers recognized for foundational contributions rather than executive roles. He has testified before the U.S. Senate on AI safety and capabilities — bringing the technical perspective of a researcher who understands, at a mathematical level, how AI systems grow in power.
Net Worth
Forbes estimated Kaplan’s net worth at approximately $3.7 billion as of early 2026, reflecting his co-founder equity in Anthropic at the company’s current valuation. If Anthropic proceeds with its targeted IPO in late 2026, this figure could change substantially.
Frequently Asked Questions
What is Jared Kaplan known for?
Jared Kaplan is best known for co-discovering AI scaling laws — the mathematical relationships that predict how AI model performance improves with more compute, data, and parameters. His 2020 paper “Scaling Laws for Neural Language Models” is foundational to modern AI development.
What is Jared Kaplan’s role at Anthropic?
Kaplan is the Chief Science Officer of Anthropic, responsible for the company’s scientific research direction and the technical foundations of Claude’s development.
What is Jared Kaplan’s net worth?
Forbes estimated Jared Kaplan’s net worth at approximately $3.7 billion as of early 2026, based on his co-founder equity stake in Anthropic.
Current flagship: Claude Opus 4.7 (claude-opus-4-7). Current models: Opus 4.7 · Sonnet 4.6 · Haiku 4.5. Claude Opus 4.7 (claude-opus-4-7) is the current flagship as of April 16, 2026. Where this article references Opus 4.6 or earlier models, those references are historical. See current model tracker →. See current model tracker →
Claude AI · Fitted Claude
Claude AI is a family of large language models built by Anthropic, a San Francisco-based AI safety company. In 2026, Claude competes directly with ChatGPT, Gemini, and Grok — and in many professional use cases, it outperforms all of them. This guide covers what Claude is, how it works, what it costs, and how to start using it today.
What Is Claude AI?
Claude is an AI assistant developed by Anthropic, a company founded in 2021 by former OpenAI researchers including Dario Amodei, Daniela Amodei, and five other co-founders. The name “Claude” is a nod to Claude Shannon, the father of information theory.
Unlike some AI tools built primarily for speed or image generation, Claude was designed from the ground up with safety and helpfulness as co-equal priorities. Anthropic uses a technique called Constitutional AI — a method of training models to follow a set of principles rather than just optimize for user approval. The result is an assistant that tends to be more careful, more honest, and less likely to hallucinate than its competitors.
As of April 2026, Claude is available through:
Claude.ai — the web and mobile interface (free and paid plans)
Claude desktop app — native Mac and Windows applications
Claude API — for developers building AI-powered applications
Claude Code — a terminal-native AI coding tool
Enterprise deployments — via Anthropic’s enterprise and team offerings
Which Claude Models Exist in 2026?
Anthropic currently offers three tiers of Claude models, each optimized for different use cases:
Model
Best For
Context Window
Notable Benchmark
Claude Opus 4.7
Complex reasoning, research, coding
200K tokens
80.8% SWE-bench, 91.3% GPQA Diamond
Claude Sonnet 4.6
Everyday tasks, balanced performance
200K tokens
Best speed-to-intelligence ratio
Claude Haiku 4.5
Fast, lightweight tasks
200K tokens
Fastest response time
All models support a 200,000-token context window by default — roughly 150,000 words, or an entire novel. Enterprise customers can access up to 500,000 tokens, and Claude Code extends to 1 million tokens for large codebase analysis.
How Does Claude AI Work?
Claude is a large language model (LLM) — a type of neural network trained on vast amounts of text data to predict and generate human-like responses. What distinguishes Claude from other LLMs is Anthropic’s emphasis on alignment and safety during training.
Claude uses two key training innovations:
Constitutional AI (CAI): Instead of relying solely on human feedback to shape model behavior, Anthropic trains Claude to evaluate its own outputs against a set of written principles. This makes Claude more consistent in avoiding harmful outputs, even in edge cases human reviewers might not anticipate.
RLHF (Reinforcement Learning from Human Feedback): Human trainers rate Claude’s responses, and those ratings guide the model toward more helpful, accurate, and appropriate answers over time.
The combination produces a model that tends to acknowledge uncertainty, push back on false premises, and decline harmful requests more gracefully than many competitors.
What Can Claude AI Do?
Claude’s capabilities in 2026 span well beyond simple chatting. Here’s what it handles well:
Writing and Editing
Claude excels at long-form content: blog posts, essays, reports, marketing copy, email sequences, legal documents, and fiction. Its writing is notably less robotic than many AI tools, partly because it’s trained to match tone and style from context clues.
Coding and Software Development
Claude Code — Anthropic’s terminal-native coding tool — has become one of the most popular AI coding environments among professional developers. It can write, debug, refactor, and explain code across virtually all major programming languages, and it understands large codebases through its million-token context window.
Research and Analysis
Claude reads and synthesizes PDFs, research papers, financial reports, and legal filings. With 200K tokens of context, it can process an entire book-length document and answer specific questions about it.
Data Analysis
Claude can read CSV files, interpret charts, write Python or SQL to analyze datasets, and explain findings in plain language — making it useful for anyone who works with data but isn’t a dedicated data scientist.
Multimodal Inputs
Claude accepts text, images, PDFs, and documents as inputs. It can describe images, extract text from screenshots, and analyze visual data — though it cannot generate images itself (for image generation, tools like Midjourney or DALL-E are required).
Claude AI Pricing: Free vs. Paid Plans in 2026
Anthropic offers four main tiers for individual users:
Plan
Price
What You Get
Best For
Free
$0/month
Limited daily messages, Claude Sonnet 4.6 access
Casual or occasional use
Claude Pro
$20/month
5x more usage, priority access, Projects
Regular users, professionals
Claude Max 5x
$100/month
5x Pro usage, Claude Code access, extended thinking
Power users, developers
Claude Max 20x
$200/month
20x Pro usage, highest priority
Heavy professional use
Enterprise plans are available with custom pricing, SSO, admin controls, extended context (up to 500K tokens), and zero-data-retention options for sensitive industries.
Claude vs. ChatGPT: What’s the Difference?
This is the question most people ask when they first hear about Claude. The honest answer: they’re both capable, and the best choice depends on your use case.
Factor
Claude
ChatGPT
Best at
Long documents, nuanced writing, coding
General tasks, image generation, plugins
Context window
200K tokens (standard)
128K tokens (GPT-4o)
Image generation
No (analysis only)
Yes (DALL-E integration)
Safety emphasis
Very high (Constitutional AI)
High
Code quality
Among the best (SWE-bench leader)
Strong
Price
$20-$200/month
$20/month (Plus), $200/month (Pro)
For most professional writing, legal/financial analysis, and software development tasks, Claude holds a measurable edge. For tasks requiring image generation or deep integration with third-party plugins, ChatGPT’s ecosystem is broader.
How to Get Started with Claude AI
Getting started takes about two minutes:
Go to claude.ai and create a free account with your email or Google login.
Start a new conversation. Type or paste your first prompt.
If you need to analyze a document, click the paperclip icon to upload PDFs, images, or files.
For power use, upgrade to Claude Pro for Projects — a feature that lets you create persistent knowledge bases that Claude remembers across conversations.
Spinning Up the API?
I can walk you through setup, model selection, and cost management — before you burn credits figuring it out yourself.
If you’re a developer, visit console.anthropic.com to get your API key and explore the Claude API.
Claude AI: Key Limitations to Know
No tool is perfect. Here are Claude’s genuine limitations as of 2026:
No image generation: Claude cannot create images. For that, you need a dedicated tool like Midjourney, DALL-E, or Stable Diffusion.
Rate limits on free and Pro plans: Heavy users — especially on the Pro tier — regularly hit daily message limits. This is the most common complaint among power users. The Max plans ($100/$200/month) solve this for most use cases.
No real-time web access by default: Unless explicitly connected to a web search tool, Claude’s knowledge has a training cutoff. It cannot browse the web in real time by default on the consumer interface.
Occasional refusals: Claude’s safety training sometimes makes it overly cautious on topics that are legitimate but touch sensitive areas. This has improved substantially with each model generation.
Frequently Asked Questions About Claude AI
Is Claude AI free?
Yes — Claude has a free tier that gives you limited daily access to Claude Sonnet 4.6. The free tier is useful for casual use, but heavy users will quickly encounter rate limits. Paid plans start at $20/month.
Who made Claude AI?
Claude was created by Anthropic, an AI safety company founded in 2021. Anthropic was started by seven former OpenAI researchers, including CEO Dario Amodei and President Daniela Amodei.
Is Claude AI better than ChatGPT?
It depends on the task. Claude generally outperforms ChatGPT on coding benchmarks, long-document analysis, and nuanced writing. ChatGPT has a broader plugin ecosystem and native image generation. Many professionals use both.
Does Claude store my conversations?
By default, Anthropic may use conversations from consumer accounts to improve its models (you can opt out in settings). Business and API customers can access zero-data-retention options. Conversation data is retained for up to five years unless you delete it manually.
Can Claude generate images?
No. Claude can analyze and describe images, but it cannot generate them. For AI image creation, use Midjourney, DALL-E, or Adobe Firefly.
What is Claude’s context window?
Standard Claude models have a 200,000-token context window — roughly 150,000 words. Enterprise plans extend this to 500,000 tokens. Claude Code supports up to 1 million tokens for large codebase analysis.
How do I access Claude Code?
Claude Code is available as part of the Claude Max subscription ($100+/month) or via the Anthropic API. It runs as a terminal-native tool — install it with npm install -g @anthropic-ai/claude-code and authenticate with your API key.
This guide is updated regularly as Anthropic ships new models and features. Last updated: April 2026.
Daniela Amodei is the President and co-founder of Anthropic, the AI safety company behind Claude. While her brother Dario Amodei serves as CEO and is the more publicly visible figure, Daniela runs the operational, commercial, and go-to-market sides of one of the most consequential AI companies in the world. She is, in practical terms, the reason Anthropic functions as a business.
Quick facts: Daniela Amodei — President and co-founder of Anthropic. Previously VP of Operations at OpenAI. Before that: Stripe, Ropes & Gray. Co-founded Anthropic in 2021 with her brother Dario and five other former OpenAI researchers. Responsible for Anthropic’s business operations, sales, partnerships, and go-to-market strategy.
Who Is Daniela Amodei?
Daniela Amodei is the President of Anthropic, the AI safety company she co-founded in 2021 alongside her brother Dario Amodei and a group of senior researchers who departed OpenAI together. While Dario leads research and product as CEO, Daniela leads everything that keeps the company running as a viable business: revenue, partnerships, hiring, operations, and the commercial strategy behind Claude.
She is among the most powerful operators in the AI industry — not a figurehead co-founder, but the executive who built Anthropic’s commercial foundation from zero while the research team focused on the models.
Background and Career Before Anthropic
Before Anthropic, Daniela spent years in operational and business roles that would prove directly relevant to building a fast-moving AI company from scratch.
She attended Dartmouth College, where she studied economics. Her early career included a position at Ropes & Gray, a prominent law firm, before moving into the technology sector. She joined Stripe — the payments infrastructure company — where she worked in business operations during a period of significant growth for the company.
The pivotal move came when she joined OpenAI as VP of Operations. She was one of the senior leaders who left OpenAI in 2020 and 2021 along with her brother Dario to found Anthropic. That cohort included several of OpenAI’s most senior researchers and operators, making it one of the most significant team departures in AI industry history.
Role at Anthropic
As President, Daniela’s domain at Anthropic covers the business side of the company end to end. Where Dario focuses on research direction, safety philosophy, and model development, Daniela owns:
Revenue and commercial growth — enterprise sales, partnerships, and the Claude business
Go-to-market strategy — how Anthropic positions and sells Claude to individuals, developers, and enterprises
Operations — the internal systems and processes that let a growing AI company function
Partnerships — major deals including Anthropic’s relationship with Amazon Web Services, one of the largest infrastructure commitments in AI company history
Hiring and team building — scaling the organization while maintaining culture
The division of labor between Daniela and Dario mirrors a pattern common in successful tech companies: one founder focused on product and technology, one focused on the business that makes the technology sustainable. At Anthropic, that structure is unusually clean and appears to function well.
Daniela Amodei and the Amazon Partnership
One of the most significant commercial milestones under Daniela’s leadership as President was securing Anthropic’s partnership with Amazon Web Services. Amazon committed to invest up to $4 billion in Anthropic, with Claude models made available through AWS’s Bedrock platform. This deal established Anthropic’s commercial credibility and gave it the infrastructure scale to compete with OpenAI and Google DeepMind.
Partnerships of this scale require sustained executive relationships and months of commercial negotiation — the kind of work that falls squarely in Daniela’s domain.
The Amodei Siblings Running Anthropic
The dynamic between Daniela and Dario Amodei at Anthropic is worth understanding because it’s unusual. Co-founders who are siblings and who have distinct, non-overlapping domains are relatively rare. In most tech companies, co-founders compete for influence. At Anthropic, the operational split appears deliberate and functional: Dario owns the mission and the models, Daniela owns the machine that funds the mission.
Dario has spoken publicly about AI safety, the risks of powerful AI systems, and Anthropic’s research philosophy. Daniela tends to operate more quietly — she is less frequently the face of Anthropic in press interviews but is consistently present in the company’s major commercial announcements and partnership moments.
Net Worth and Anthropic’s Valuation
Anthropic has raised billions of dollars in venture funding from investors including Google, Amazon, and Spark Capital, with valuations that have grown significantly through each funding round. As a co-founder and President holding equity in the company, Daniela Amodei’s net worth is tied primarily to Anthropic’s private valuation.
Anthropic is not publicly traded, so precise figures are not available. At the company’s reported valuations, co-founders with meaningful equity stakes hold substantial paper wealth — though the actual liquidity of that wealth depends on if and when Anthropic conducts an IPO or secondary transactions.
Why Daniela Amodei Matters for Claude
Claude exists because Anthropic exists as a viable company. Daniela Amodei is one of the primary reasons Anthropic is viable. The research team can build frontier AI models, but without a functioning commercial operation those models don’t reach users, don’t generate revenue, and don’t fund the next generation of research.
Every enterprise Claude deployment, every API integration, every AWS customer using Claude through Bedrock, every API integration, every AWS customer using Claude through Bedrock — these exist in part because of the commercial infrastructure Daniela has built. The Claude you use is as much a product of her work as it is of the research team’s.
Frequently Asked Questions
Who is Daniela Amodei?
Daniela Amodei is the President and co-founder of Anthropic, the AI company behind Claude. She previously served as VP of Operations at OpenAI before co-founding Anthropic in 2021 with her brother Dario Amodei and other former OpenAI researchers.
Is Daniela Amodei related to Dario Amodei?
Yes. Daniela and Dario Amodei are siblings. Dario is the CEO of Anthropic; Daniela is the President. They co-founded Anthropic together in 2021 along with five other former OpenAI researchers.
What does Daniela Amodei do at Anthropic?
As President, Daniela oversees Anthropic’s business operations, commercial strategy, revenue, partnerships, and go-to-market. She is responsible for the business side of Anthropic while Dario leads research and product.
Where did Daniela Amodei work before Anthropic?
Before co-founding Anthropic, Daniela was VP of Operations at OpenAI. Prior to OpenAI she worked at Stripe in business operations, and earlier in her career she was at the law firm Ropes & Gray. She studied economics at Dartmouth College.
What is Daniela Amodei’s net worth?
Daniela Amodei’s net worth is not publicly known — Anthropic is a private company and does not disclose individual equity stakes. Her net worth is tied primarily to her equity in Anthropic, which has been valued at billions of dollars across successive funding rounds from investors including Amazon and Google.
Anthropic introduced Dreaming at Code w/ Claude (May 6, 2026) — a new Managed Agents capability where agents review their own session history overnight to improve future performance. Harvey (legal AI) reported a roughly 6× task completion rate increase after implementing it. Dreaming is developer-access preview only. Multiagent Orchestration and Outcomes are now in public beta. See the new Dreaming section below.
What Is Claude Managed Agents? (Current Status, May 2026)
Claude Managed Agents is Anthropic’s framework for long-running, stateful AI agents — agents that can maintain context across sessions, hand off between sub-agents, and now, improve themselves by reviewing their own work history. Here’s the current status of each component:
Component
Status
Who Has Access
Multiagent Orchestration
Public Beta
All API developers
Outcomes
Public Beta
All API developers
Dreaming
Developer Preview
Selected developers only
Dreaming: The Feature the Press Mostly Missed
Announced at Code w/ Claude on May 6, 2026, Dreaming is a Managed Agents capability that lets agents review and reorganize their own memory between sessions. The mechanism:
After a session ends, the agent reads its existing memory store alongside the session transcripts
It produces a new, reorganized memory store: duplicates merged, stale entries replaced, new patterns surfaced
The next session starts with a higher-quality knowledge base — capturing insights no single session could hold
This is meaningfully different from simply persisting conversation history. The agent isn’t just remembering what happened — it’s synthesizing what it learned. Think of it as the difference between taking notes and actually reviewing and reorganizing your notes the next morning.
The Harvey Result
Harvey, the legal AI company, reported approximately a 6× task completion rate increase after implementing Dreaming in their Managed Agents workflow. Harvey’s use case — complex legal research that spans multiple sessions with evolving context — is exactly the kind of work Dreaming was designed for. Sessions build on each other rather than starting fresh each time.
Dreaming is not available to end users — it’s a developer-layer capability requiring implementation
It’s not persistent memory in the claude.ai chat interface
It’s not available to free or standard Pro subscribers through any interface
It’s a developer preview, not GA — expect it to evolve before full release
Our Take: Why This Architecture Matters
We run Managed Agents in our own Cowork workflows. The Dreaming announcement is the first time Anthropic has shipped something that resembles how expert human knowledge actually compounds over time — not by accumulating raw notes, but by periodically synthesizing and reorganizing what’s been learned into a cleaner structure.
The Harvey 6× result is a real-world data point from a production legal AI workflow. That’s not a benchmark number — it’s a deployed system showing measurable improvement from session-to-session memory refinement. Whether that 6× figure holds across different use cases is unknown, but the direction of the effect is the signal: agents that learn from their own history outperform agents that don’t.
For non-developer users watching this space: Dreaming is the preview of what agentic AI will look like when it becomes mainstream. The groundwork being laid now in developer preview will eventually surface in subscription-tier products.
Model Accuracy Note — Updated May 2026
Current flagship: Claude Opus 4.7 (claude-opus-4-7). Current models: Opus 4.7 · Sonnet 4.6 · Haiku 4.5. Claude Opus 4.7 (claude-opus-4-7) is the current flagship as of April 16, 2026. Where this article references Opus 4.6 or earlier models, those references are historical. See current model tracker →. See current model tracker →
Tygart Media Strategy
Volume Ⅰ · Issue 04Quarterly Position
By Will Tygart • Long-form Position • Practitioner-grade
You opened this tab because you need a number you can actually use. Not a vibe, not “it depends.” A real pricing breakdown you can put in a spreadsheet, a budget request, or a Slack message to your CTO.
This is that page. Every pricing variable for Claude Managed Agents in one place, verified against Anthropic’s current documentation as of April 2026. Bookmark it. The beta will update; so will this.
Quick Reference: The Formula
Total Cost = Token Costs + Session Runtime ($0.08/hr) + Optional Tools
Session runtime only accrues while status = running. Idle time is free.
The Two Cost Dimensions
Claude Managed Agents bills on exactly two dimensions: tokens and session runtime. Every pricing question you have collapses into one of these two buckets.
Dimension 1: Token Costs
These are identical to standard Claude API pricing. You pay the same rates you’d pay calling the Messages API directly. No Managed Agents markup on tokens. Current rates for the models most commonly used in agent work:
Claude Sonnet 4.6: ~$3/million input tokens, ~$15/million output tokens
Prompt caching: same multipliers as standard API — cache hits dramatically reduce input token costs on long sessions with stable system prompts
The implication: a token-heavy agent with a large system prompt that runs the same context repeatedly benefits significantly from prompt caching, and that benefit carries over unchanged into Managed Agents.
Dimension 2: Session Runtime — $0.08/Session-Hour
This is the Managed Agents-specific charge. You pay $0.08 per hour of active session runtime, metered to the millisecond.
The critical word is active. Runtime only accrues while your session’s status is running. The following do not count toward your bill:
Time spent waiting for your next message
Time waiting for a tool confirmation
Idle time between tasks
Rescheduling delays
Terminated session time
This is not how you’d bill a virtual machine. It’s closer to how AWS Lambda bills — you pay for execution, not reservation. An agent that “runs” for 8 hours but spends 6 of those hours waiting on human input has a very different bill than one running continuous autonomous loops.
Optional Tool Costs
Web Search: $10 per 1,000 Searches
If your agent uses web search, each search costs $10/1,000 — that’s $0.01 per search. For most agents, this is negligible. For a research agent running hundreds of searches per session, it becomes a line item worth modeling separately.
Code Execution: Included in Session Runtime
Code execution containers are included in your $0.08/session-hour charge. You’re not separately billed for container hours on top of session runtime. This is explicitly stated in Anthropic’s docs and represents meaningful savings versus provisioning your own compute.
Worked Cost Examples
Example 1: Daily Research Agent
Runs once per day. 30 minutes of active execution. Processes 10 documents, outputs a summary report. Moderate token volume.
In practice, no agent has zero idle time — real cost will be lower
Token costs at this scale become the dominant factor by a wide margin
Anthropic’s Official Example (from their docs)
A one-hour coding session using Claude Opus 4.7 consuming 50,000 input tokens and 15,000 output tokens: session runtime = $0.08. With prompt caching active and 40,000 of those tokens as cache reads, the token costs drop significantly. The runtime charge stays flat at $0.08 regardless of caching.
What’s Not Billed in Managed Agents
A few things that might seem like costs but aren’t:
Infrastructure provisioning: Anthropic handles hosting, scaling, and monitoring at no additional charge
Container hours: Explicitly not separately billed on top of session runtime
State management and checkpointing: Included in the session runtime charge
Error recovery and retry logic: Anthropic’s infrastructure problem, not yours
Rate Limits
Managed Agents has specific rate limits separate from standard API limits:
Create endpoints: 60 requests/minute
Read endpoints: 600 requests/minute
Organization-level limits still apply
For higher limits, contact Anthropic enterprise sales
How to Access Managed Agents Pricing
Managed Agents is available to all Anthropic API accounts in public beta. No separate signup, no premium tier gate. You need the managed-agents-2026-04-01 beta header in your API requests — the Claude SDK adds this automatically.
For high-volume agent applications, Anthropic’s enterprise sales team negotiates custom pricing arrangements. Contact them at [email protected] or through the Claude Console.
The Pricing Signals Worth Noting
Anthropic recently ended Claude subscription access (Pro/Max) for third-party agent frameworks, requiring those users to switch to pay-as-you-go API pricing. This signals a deliberate strategy: consumer subscriptions are for human-paced interactions; agent workloads route through the API. The $0.08/session-hour rate exists in that context — it’s infrastructure pricing for compute that runs beyond human attention spans.
The session-hour model also signals something about Anthropic’s infrastructure cost structure. They’re pricing on active execution time because that’s what actually taxes their systems. Idle sessions don’t cost them much; active agents do. The billing model follows the actual resource consumption pattern.
Frequently Asked Questions
Is the $0.08/session-hour charge in addition to token costs, or does it replace them?
In addition to. You pay both: standard token rates for all input and output tokens, plus $0.08 per hour of active session runtime. They’re separate line items.
Does prompt caching work in Managed Agents sessions?
Yes. Prompt caching multipliers apply identically to Managed Agents sessions as they do to standard API calls. If your agent has a large, stable system prompt, caching it can significantly reduce input token costs.
What happens if my session crashes? Am I billed for the crashed time?
Runtime accrues only while status is running. Terminated sessions stop accruing. Anthropic’s infrastructure handles checkpointing and crash recovery — the session state is preserved even if the session terminates unexpectedly.
Can I use Managed Agents on the free API tier?
Managed Agents is available to all Anthropic API accounts in public beta, but standard tier access and rate limits apply. Free API tier users receive a small credit for testing.
How does this compare to running agents on my own infrastructure?
