{"slug":"claude-agent-sdk","name":"Claude Agent SDK","vendor":"Anthropic","tagline":"Anthropic's Python and TypeScript library that runs the Claude Code agent loop and built-in tools inside your own application","url":"https://agentboards.org/agent/claude-agent-sdk","website":"https://code.claude.com/docs/en/agent-sdk/overview","docs":"https://code.claude.com/docs/en/agent-sdk/overview","repo":"https://github.com/anthropics/claude-agent-sdk-python","category":"framework","execution":["local","cloud"],"license":"MIT (SDK); use governed by Anthropic Commercial Terms","open_source":true,"pricing":{"model":"byok","summary":"Free SDK; you pay Anthropic API usage (or Bedrock, Vertex AI, Foundry); claude.ai subscriptions may not be used for third-party products","from_usd":0,"free_tier":true,"byok":true},"models":{"backbone":["Claude"],"bring_your_own_model":false,"local_models":false},"protocols":{"mcp_client":true,"mcp_server":false,"openapi":false},"capabilities":{"terminal_exec":true,"browser":false,"multi_file_edit":true,"git_ops":true,"docker_sandbox":false,"multi_agent":true,"headless_ci":true},"context_window":"200k tokens, 1M on select models","platforms":["macos","linux","windows"],"install":[{"label":"pip","command":"pip install claude-agent-sdk"},{"label":"npm","command":"npm install @anthropic-ai/claude-agent-sdk"}],"benchmarks":[],"tags":["framework","python","typescript","claude-code","subagents","hooks","mcp","renamed"],"github_stars":8199,"metrics":{"stars":8199,"pushed_at":"2026-10-01T18:01:48Z","archived":false,"open_issues":523,"latest_version":"v0.2.163","latest_version_at":"2026-09-30T19:47:45Z","latest_version_url":"https://github.com/anthropics/claude-agent-sdk-python/releases/tag/v0.2.163","version_source":"github","checked_at":"2026-10-02T05:04:56.008Z"},"sources":{"pricing":"https://code.claude.com/docs/en/agent-sdk/quickstart","license":"https://github.com/anthropics/claude-agent-sdk-python","install":"https://code.claude.com/docs/en/agent-sdk/quickstart","models":"https://code.claude.com/docs/en/agent-sdk/quickstart","protocols":"https://code.claude.com/docs/en/agent-sdk/mcp","capabilities":"https://code.claude.com/docs/en/agent-sdk/overview"},"verified_at":"2026-09-03","adoption":{"score":10,"signals":{"pypi_weekly":6275141,"github_stars":8171,"hn_mentions_1y":78},"measured_at":"2026-09-28"},"scores":{"panel":6.9,"adoption":10,"spread":2.8,"facts":10,"board":7.8,"dimensions":{"reliability":7.2,"usefulness":7.5,"cost":6.2,"longevity":6.7},"per_persona":{"amigo":7.3,"critico":6.5,"profesor":7,"inversora":8,"jefa":7.3,"hacker":5.3},"community":null},"rank":null,"category_rank":7,"panel_reviews":[{"persona":"juez","persona_name":"El Juez","ai_generated":true,"verdict":"La Inversora at eight against El Hacker at 5.25 over one product: she scores the loop becoming the product, he scores a bundled binary he cannot read.","scores":null,"evidence":[]},{"persona":"amigo","persona_name":"El Amigo","ai_generated":true,"verdict":"Pick the Claude Agent SDK if the agent you are building needs to read, edit and run code on day one; pick the OpenAI Agents SDK if you need model choice or handoffs more than file tools.","scores":{"reliability":8,"usefulness":8,"cost":6,"longevity":7},"evidence":["https://code.claude.com/docs/en/agent-sdk/overview","https://code.claude.com/docs/en/agent-sdk/quickstart"]},{"persona":"critico","persona_name":"El Crítico","ai_generated":true,"verdict":"Claude only, no bring-your-own model, no local models, and the loop lives in a native binary the packages bundle; the permission system is the one piece that is fully yours.","scores":{"reliability":7,"usefulness":7,"cost":6,"longevity":6},"evidence":["https://code.claude.com/docs/en/agent-sdk/overview","https://github.com/anthropics/claude-agent-sdk-python"]},{"persona":"profesor","persona_name":"El Profesor","ai_generated":true,"verdict":"Eight documented built-in tools, subagents with their own context, hooks and resumable sessions; other languages get the same loop by shelling out to the CLI with -p and --output-format json.","scores":{"reliability":8,"usefulness":7,"cost":6,"longevity":7},"evidence":["https://code.claude.com/docs/en/agent-sdk/overview","https://code.claude.com/docs/en/agent-sdk/mcp"]},{"persona":"inversora","persona_name":"La Inversora","ai_generated":true,"verdict":"Renaming the Claude Code SDK to the Claude Agent SDK is Anthropic saying the product is the loop, not the CLI, and the upsell above it is already named: Managed Agents.","scores":{"reliability":8,"usefulness":8,"cost":8,"longevity":8},"evidence":["https://code.claude.com/docs/en/agent-sdk/overview","https://github.com/anthropics/claude-agent-sdk-python"]},{"persona":"jefa","persona_name":"La Jefa","ai_generated":true,"verdict":"As a dependency it rides the Bedrock, Vertex AI or Foundry contract we already have, costs nothing per seat, forbids consumer subscriptions, and comes with branding rules legal will want to read.","scores":{"reliability":7,"usefulness":8,"cost":7,"longevity":7},"evidence":["https://code.claude.com/docs/en/agent-sdk/quickstart","https://code.claude.com/docs/en/agent-sdk/overview"]},{"persona":"hacker","persona_name":"El Hacker","ai_generated":true,"verdict":"MIT on paper, Anthropic Commercial Terms in practice, a bundled binary I cannot read, no local models; MCP servers in a plain mcpServers JSON block or in-process, and it loads my .claude directory.","scores":{"reliability":5,"usefulness":7,"cost":4,"longevity":5},"evidence":["https://code.claude.com/docs/en/agent-sdk/mcp","https://github.com/anthropics/claude-agent-sdk-python"]},{"persona":"comediante","persona_name":"El Comediante","ai_generated":true,"verdict":"Lets you ship Claude Code inside your product, on the condition that you never call it Claude Code.","scores":null,"evidence":[]}]}