A catalog that compounds. Companies add tools via GitHub PR. Developers find them via semantic search or directly inside Claude. Every tool added makes the whole system more useful.
Open a GitHub pull request with a YAML file. No forms, no waiting for manual review cycles. The PR validator checks your file automatically, and on merge the tool is embedded and live within minutes.
tools/agents/your-tool.yaml → /tools/your-toolThe sync pipeline generates a 1536-dimension vector embedding from your tool's description, problem statement, and use cases using OpenAI's text-embedding-3-small model. This is what makes "find me a vector database for Python with a free tier" work.
pgvector cosine similarity · threshold 0.2 · instant on mergeSemantic search means developers don't need to know your tool's name. They describe a problem — "add observability to my LangChain app" or "fine-tune a model on my own data" — and your tool surfaces based on relevance to the problem it actually solves.
agenticaiforgood.com/toolsThe MCP server gives Claude direct access to the full catalog. Developers ask their AI assistant to recommend tools for their stack without leaving their editor. Your tool is findable from inside every Claude Code, Claude Desktop, and any MCP-compatible AI coding setup.
npx -y agentic-ai-for-good-mcpEach tool page shows copy-ready integration snippets for Claude Code, Codex, Cursor, and Windsurf — so developers go from finding your tool to using it inside their agentic setup immediately. Install command, CLAUDE.md entry, and use case prompts included.
Claude Code · Codex · Cursor · WindsurfEvery new tool makes the semantic search smarter. Every developer who finds a tool and uses it becomes a potential contributor who adds the next one. The catalog compounds — the more tools there are, the more useful every query becomes.
The flywheelThis is everything you need to add a tool. The more specific your problem_solved and use_cases fields, the better the semantic search quality.
name: Your Tool Name description: 2-3 sentences on what it does. github_url: https://github.com/org/your-tool website_url: https://yourtool.com install_command: pip install your-tool category: Agents # LLM | Vector DB | RAG | Agents | Fine-tuning | Monitoring | Data | Dev Tools tags: [python, llm, agents] pricing: free # free | freemium | paid is_open_source: true # This powers semantic search — be specific problem_solved: | What was painful before this tool? What can developers stop doing manually? The more concrete, the better. # At least 3. These improve search quality dramatically. use_cases: - Build a RAG pipeline that answers questions from internal docs - Create an autonomous agent that browses and summarizes the web - Orchestrate multi-step LLM workflows without boilerplate