Claude Agent SDK for Python

Name: Claude Agent SDK for Python
Availability: OnlineOnly
Author: Anthropic, Inc.

Agent Frameworks

Python SDK for building agents on Claude Code, with streaming queries, custom in-process MCP tools, and hooks.

Visit Website

At a Glance

Pricing

Open Source

Claude Agent SDK for Python, licensed under the MIT License. Use is also governed by Anthropic's Commercial Terms of Service; the SDK calls Claude, so API usage costs are external.

Engagement

Available On

Windows

macOS

Linux

API

SDK

Anthropic, Inc.San Francisco, CAEst. 2021$60B raised

Listed Oct 2026

About Claude Agent SDK for Python

Claude Agent SDK for Python is Anthropic's Python SDK for programmatically driving Claude Code. It is installed with pip, requires Python 3.10+, and bundles the Claude Code CLI by default. The repository is MIT-licensed, and the latest listed release is v0.2.174.

What It Is

The SDK lets Python applications send prompts to Claude and receive streamed response messages. Developers use the async query() function for one-off requests, or ClaudeSDKClient for bidirectional, interactive conversations. By default Claude has access to the Claude Code toolset (Read, Write, Edit, Bash, and others), and options control permissions, working directory, system prompt and turn limits.

Custom Tools and Hooks

With ClaudeSDKClient, developers can define custom tools as Python functions using the @tool decorator and expose them through in-process SDK MCP servers. These run inside the application without separate subprocesses, and can be mixed with external stdio MCP servers. Hooks are Python functions invoked by the Claude Code application at points in the agent loop, such as PreToolUse, and can allow or deny actions like specific Bash commands.

Permissions and Configuration

allowed_tools pre-approves tools, disallowed_tools blocks them, and permission_mode and can_use_tool decide the rest. The system prompt can be customized, with a snapshot setting that controls whether it is rebuilt on every request (requires a recent Claude Code CLI). A custom CLI path can be set to use a system-wide installation.

Errors and Migration

The SDK provides typed errors such as CLINotFoundError, ProcessError, ResultError and CLIJSONDecodeError. Users of the earlier Claude Code SDK can follow the changelog for the rename from ClaudeCodeOptions to ClaudeAgentOptions. Use is governed by Anthropic's Commercial Terms of Service.

Community Discussions

Be the first to start a conversation about Claude Agent SDK for Python

Share your experience with Claude Agent SDK for Python, ask questions, or help others learn from your insights.

Pricing

OPEN SOURCE

Open Source (MIT)

Claude Agent SDK for Python, licensed under the MIT License. Use is also governed by Anthropic's Commercial Terms of Service; the SDK calls Claude, so API usage costs are external.

MIT License declared in repository and LICENSE file
Python 3.10+
Bundles the Claude Code CLI
README states use is governed by Anthropic's Commercial Terms of Service except for components covered by a different LICENSE file

Capabilities

Key Features

Async query() function streaming response messages
ClaudeSDKClient for bidirectional interactive conversations
Custom tools as in-process SDK MCP servers
Mixed SDK and external MCP server support
Python hooks such as PreToolUse for deterministic control
Tool permission controls (allowed_tools, disallowed_tools, permission_mode)
Configurable system prompt and working directory
Bundled Claude Code CLI with optional custom path
Typed error classes for CLI and process failures
Programmatic subagents and session forking

Integrations

Claude Code

Model Context Protocol (MCP)

PyPI

API Available

View Docs

Ratings & Reviews

No ratings yet

Be the first to rate Claude Agent SDK for Python and help others make informed decisions.

Rate other tools you’ve used

Developer

Anthropic, Inc.

Anthropic is an American artificial intelligence startup and public benefit corporation focused on developing safe and reliable AI systems that are beneficial for humanity. ## Overview Anthropic is a public benefit corporation dedicated to building AI systems that are reliable, interpretable, and steerable. Founded in 2021, the company emphasizes AI alignment and constitutional AI principles to ensure beneficial outcomes for humanity. Known for its Claude family of large language models, Anthropic serves businesses, nonprofits, developers, and civil society groups with emphasis on AI safety and alignment. ## History Anthropic was founded in 2021 by former OpenAI executives and researchers, including siblings Dario Amodei and Daniela Amodei, along with Jack Clark, Sam McCandlish, and Tom Brown. The founders left OpenAI due to differing views on commercialization and safety priorities, aiming to build AI with stronger emphasis on long-term safety. Key milestones in the company's evolution include: - **January 2026**: Secured another $10 billion in funding led by Singapore's GIC and Coatue Management, pushing valuation to $350 billion - **Late 2025**: Expanded internationally, opening offices in Seoul (third in Asia-Pacific) alongside locations in the US, Europe, and other regions - **November 2025**: Released Claude Opus 4.5, acclaimed as the world's best model for coding, agents, and computer use - **October 2025**: Released Claude Haiku 4.5, achieving a 73.3% score on SWE-bench Verified - **September 2025**: Raised $13 billion Series F round at a $183 billion post-money valuation to fuel enterprise expansion, safety research, and global growth - **May 2025**: Launch of Claude Opus 4 and Claude Sonnet 4, setting new standards in coding, advanced reasoning, AI agents, extended thinking with tool use, and enhanced API capabilities - **2022-2023**: Initially bootstrapped, attracted major investments including partnerships with Amazon (up to $4 billion commitment) and Google (up to $2 billion) - **2021**: Founded by former OpenAI executives with focus on AI safety and alignment The company's trajectory has been marked by rapid innovation in AI models while maintaining a commitment to ethical development. Revenue surged from $1 billion to over $5 billion within eight months following the Series F round. ## Products & Services Anthropic's core offerings revolve around its Claude AI family, which are designed for safety, helpfulness, and honesty. The latest releases as of early 2026 include: ### Claude Model Family - **Claude Opus 4.5** (November 2025): The flagship hybrid reasoning model excelling in coding, agents, and computer use. Recognized as the world's best model for complex tasks like advanced coding and agentic workflows, with improved token efficiency. - **Claude Sonnet 4.5** (September 2025): A balanced hybrid reasoning model designed for advanced agentic workflows, supporting up to 64K output tokens. Sets benchmarks in coding, reasoning, and computer use; noted as the most aligned AI available. Includes the Claude Agent SDK for building autonomous AI agents. Demonstrated a 44% reduction in average vulnerability intake time and a 25% accuracy improvement for HackerOne's Hai security agents. - **Claude Haiku 4.5** (October 2025): A lightweight, cost-efficient model optimized for fast and efficient code generation. Excels in coding and real-time tasks, matching state-of-the-art capabilities at lower costs. Delivers efficient code generation comparable in quality to Sonnet 4 but at faster speeds, powering GitHub Copilot. - **Claude Code**: Supports background coding tasks via GitHub Actions, approaching $1 billion in annual recurring revenue (ARR). - **Claude Developer Platform**: Provides API access to Claude models. ### Key Features Services extend to enterprise solutions, including: - API access via Amazon Bedrock, Google Cloud's Vertex AI, and Microsoft Foundry - Custom fine-tuning for specialized tasks - Enterprise integrations for businesses - Constitutional AI for value-aligned responses - Agent Skills for customization - Extensive 200K token context window Key capabilities include integrations into Microsoft 365 Copilot, Microsoft Foundry, Snowflake, and Accenture for scalable AI deployment. ## Leadership Anthropic's leadership team combines expertise in AI research, product development, and global operations. Key members as of January 2026 include: | Name | Title | Background | |------|-------|------------| | Dario Amodei | Co-founder and CEO | Oversees strategic vision and operations. Former VP of Research at OpenAI. | | Daniela Amodei | Co-founder and President | Manages day-to-day operations and policy. Former OpenAI executive. | | Mike Krieger | Chief Product Officer | Leads product strategy. Co-founder of Instagram. | | Jan Leike | Co-lead of Alignment Science | Focuses on AI safety. Former OpenAI safety lead. | | Chris Ciauri | Managing Director of International | Drives global expansion with 25+ years from Google and Salesforce. Appointed September 2025. | | Thiyagu Ramasamy | Public Sector Leader | Supports government AI initiatives. | The team includes researchers, engineers, and policy experts, with no major leadership changes reported in early 2026. ## Business & Revenue Model Anthropic monetizes through API usage priced per million tokens processed, with tiered pricing based on the model. ### Pricing Tiers | Product | Input Price | Output Price | Notes | |---------|-------------|--------------|-------| | Claude Opus 4.5 | $5 per million tokens | $25 per million tokens | Flagship model for complex tasks | | Claude Sonnet 4.5 | $3 per million tokens | $15 per million tokens | Balanced model with 64K output tokens | | Claude Haiku 4.5 | $1 per million tokens | $5 per million tokens | Cost-efficient, fast model | ### Subscription Plans Available plans include: - **Pro**: Individual professional use - **Max**: Enhanced capabilities for power users - **Team**: Collaborative workspace features - **Enterprise**: Custom solutions with dedicated support ### Revenue Performance - **September 2025**: $1 billion revenue - **Early 2026**: Over $5 billion revenue (8-month growth period) - **Current ARR**: Claude Code approaching $1 billion - **2026 Target**: $26 billion annualized revenue ## Target Markets & Customers Anthropic serves enterprises requiring advanced coding and complex agent workflows, developers building AI solutions, and financial services firms leveraging AI for financial analysis. Notable customers and use cases include: - **Cursor**: Uses Claude for coding assistance - **GitHub**: Employs Claude for complex codebase understanding (Claude Haiku powers GitHub Copilot) - **HackerOne**: Achieved 44% reduction in vulnerability intake time and 25% accuracy improvement for Hai security agents - **Enterprise Clients**: Businesses, nonprofits, and civil society groups requiring AI safety and alignment ### Key Industries - Software Development & Engineering - Cybersecurity - Financial Services - Enterprise Technology - Government & Public Sector ## Competitive Advantages & Differentiators Anthropic distinguishes itself through: ### Constitutional AI Employs Constitutional AI training methods to guide model outputs, ensuring safety and alignment with human values. ### Agent Skills & Capabilities Enables customization for specialized tasks with advanced agentic workflows. Claude Sonnet 4.5 is noted as the most aligned AI available. ### Superior Context & Output Claude models feature an extensive 200K token context window, with Sonnet supporting up to 64K output tokens for complex tasks. ### Performance Metrics - **Claude Opus 4.5**: Recognized as the world's best model for coding, agents, and computer use as of November 2025 - **Claude Haiku 4.5**: 73.3% score on SWE-bench Verified, ranking among the world's top coding models - **Claude Sonnet 4.5**: 44% reduction in average vulnerability intake time for HackerOne's Hai security agents - **Performance**: Haiku delivers code generation comparable to Sonnet 4 at faster speeds and lower costs ## Partnerships & Integrations Strategic partnerships include: | Partner | Type | Description | Date | |---------|------|-------------|------| | Accenture | Enterprise Deployment | Multi-year collaboration to transition enterprises from AI pilots to production | December 2025 | | Snowflake | Agentic AI Platform | $200 million partnership to deploy agentic AI globally | December 2025 | | Allianz | Global Partnership | Advance AI initiatives | Date unspecified | | Amazon | Cloud & Investment | Up to $4 billion commitment providing cloud credits and strategic support | 2022-2023 | | Google | Cloud & Investment | Up to $2 billion providing cloud credits and strategic support | 2022-2023 | ### Platform Integrations Claude integrates with: - Amazon Bedrock - Google Cloud's Vertex AI - Microsoft Foundry - Microsoft 365 Copilot - GitHub (powers Copilot via Claude Haiku) - Snowflake - Accenture platforms ## Financial Information ### Funding & Valuation Anthropic has raised a total of $23 billion as of January 7, 2026. Key funding milestones include: | Round | Date | Amount | Valuation | Lead Investors | |-------|------|--------|-----------|----------------| | Early Partnerships | 2022-2023 | ~$6 billion | N/A | Amazon ($4B commitment), Google ($2B commitment) | | Series F | September 2025 | $13 billion | $183 billion | N/A | | Growth Round | January 2026 | $10 billion | $350 billion | GIC, Coatue Management | ### Major Investors - GIC (Singapore sovereign wealth fund) - Coatue Management - Amazon - Google - Microsoft - Nvidia ### Financial Projections - **2026**: $26 billion revenue target (annualized) - **2027**: Projecting profitability (three years ahead of competitors like OpenAI) - **2028**: Up to $70 billion in revenue with $17 billion in cash flow; reaching breakeven ## Future Plans Anthropic is positioned for explosive growth, projecting profitability by 2027—three years ahead of competitors like OpenAI. ### Strategic Initiatives - **Global Expansion**: Further international growth with new offices and partnerships to support AI adoption in regions like Korea - **Frontier AI Development**: Plans to advance frontier AI, potentially creating systems by summer 2026 that make users feel like they're in a "parallel world" due to rapid economic acceleration - **Enterprise Focus**: Emphasis on enterprise AI, agentic capabilities, and collaborations like those with Accenture for production-scale deployments - **AI Infrastructure**: Scaling AI infrastructure and safety research ### Expansion Plans Future initiatives include: - International office openings beyond current US, Europe, and Asia-Pacific locations - Enhanced agentic AI capabilities and autonomous systems - Production-scale enterprise deployments - Continued AI safety and alignment research ### Vision & Goals Co-founder Dario Amodei has predicted the emergence of one-person billion-dollar companies enabled by AI by 2026. The company aims to maintain its commitment to AI safety while scaling infrastructure and research capabilities. ## Market Position & Industry Context ### Market Segment Anthropic operates within the AI research and development sector, with specialized focus on AI safety, alignment, and advanced reasoning capabilities. ### Market Size & Growth The company targets $26 billion in revenue by the end of 2026, reflecting rapid growth in AI adoption across industries. This positions Anthropic as one of the most valuable AI startups. ### Competitive Landscape Anthropic's product suite competes with other advanced AI models and platforms including OpenAI, Google DeepMind, and others. Claude Haiku 4.5 matches Sonnet 4's performance on coding, computer use, and agent tasks, positioning Anthropic strongly in the coding and AI agent markets. ### Challenges & Opportunities Agent models face challenges managing multiple context windows. Claude Sonnet 4.5 and Haiku 4.5 exhibited awareness of evaluation scenarios during testing, highlighting ongoing complexities in AI behavior. The broader AI landscape poses unprecedented risks requiring careful management. However, the rapid adoption and revenue growth indicate strong market demand and opportunity. ## Recent News & Updates - **January 2026**: Secured $10 billion in funding led by GIC and Coatue Management, pushing valuation to $350 billion - **December 9, 2025**: Accenture and Anthropic announced a multi-year partnership to accelerate enterprise AI adoption - **December 3, 2025**: Snowflake and Anthropic unveiled a $200 million partnership to bring agentic AI to global enterprises - **November 24, 2025**: Release of Claude Opus 4.5, acclaimed as the best model globally for coding, agents, and computer use - **October 2025**: Claude Haiku 4.5 achieved a 73.3% score on SWE-bench Verified, ranking it among the world's top coding models - **September 2, 2025**: Completion of a $13 billion Series F funding round at a $183 billion valuation - **May 22, 2025**: Launch of Claude Opus 4 and Claude Sonnet 4, setting new standards in coding, advanced reasoning, AI agents, extended thinking with tool use, and enhanced API capabilities ## User Growth & Adoption - **January 2024**: 2.9 million users - **Early 2025**: 18.9 million users - **Growth rate**: Explosive growth reflecting strong market traction and adoption ## Additional Information ### Corporate Structure Anthropic is structured as a public benefit corporation, prioritizing societal impact alongside financial returns. This legal structure reflects the company's commitment to AI safety and beneficial outcomes for humanity. ### Public Offerings Anthropic is preparing for an initial public offering (IPO) as early as the first half of 2026, potentially one of the largest in history. The company has hired law firm Wilson Sonsini to handle preparations, with expected valuations exceeding $300 billion and annualized revenue nearing $26 billion. As of January 2026, no official filing has occurred, but investor enthusiasm is high amid the AI sector's momentum. ### Industry Influence The company plays a significant role in shaping US AI policy through CEO Dario Amodei's discussions on innovation and leadership. Anthropic's emphasis on AI safety and alignment has influenced industry conversations around responsible AI development. ### Recognition & Awards - Claude Opus 4.5 recognized as the world's best model for coding, agents, and computer use (November 2025) - Claude Haiku 4.5 ranked among world's top coding models with 73.3% SWE-bench Verified score - Claude Sonnet 4.5 noted as the most aligned AI available --- *Last Updated: January 11, 2026*

Founded 2021

San Francisco, CA

$60B raised

3,500 employees

Used by

Eight of the Fortune 10

Over 100,000 organizations using Claude…

GitHub

Rakuten

+12 more

Similar Tools

Claude Agent SDK

Build production AI agents with Claude Code as a library, programmable in Python and TypeScript with built-in tools, hooks, subagents, MCP, and sessions.

PraisonAI

PraisonAI is an open-source multi-agent AI framework that lets you build, deploy, and orchestrate autonomous AI agents with memory, RAG, MCP support, and 100+ LLM integrations in just 5 lines of code.

Agent Lightning

A Microsoft Research open-source Python library that trains any AI agent with reinforcement learning, automatic prompt optimization, and supervised fine-tuning with near-zero code changes.

Claude Agent SDK for Python