Agent Grove
Agent Grove is an open-source knowledge base for AI agent engineering. It studies real systems, official engineering guidance, and small experiments, then turns them into concise frameworks, diagrams, case studies, and minimal examples.
This site is the reading layer of the repository. GitHub remains the source of truth for version history, code, and collaboration.
Start Here
Knowledge Framework Draft
| Area | Core Question |
|---|---|
| Agent-Ready Engineering Infrastructure | How do we engineer project context, intent, execution, verification, governance, and feedback so coding agents can work reliably? |
| Agent Core | How should agent loops, state, planning, execution, reflection, and handoff be designed? |
| Context Engineering | How should agents retrieve, select, compress, and update context? |
| Tools and MCP | How should tools, resources, prompts, skills, permissions, and side effects be modeled? |
| Memory | How should short-term state, long-term memory, user models, and skill memory be separated? |
| RAG and Knowledge | How should document parsing, chunking, retrieval, reranking, grounding, and permission filtering work? |
| Evals and Observability | How do we prove agent behavior is reliable and trace failures? |
| Coding Agents | How do agents understand codebases, edit code, run tests, and collaborate with reviewers? |
| Personal Agents | How do long-running assistants handle channels, memory, automation, and safety? |
| Model Gateway and Serving | How should model routing, fallback, cost, rate limits, caching, and serving be handled? |