TokenKnows — distill AI coding sessions into weekly reports, ADRs and a knowledge graph
@johnnywuj81
About TokenKnows — distill AI coding sessions into weekly reports, ADRs and a knowledge graph
Distill AI coding sessions (Claude Code / Codex / Cursor) into weekly reports, ADRs, incident reviews and a knowledge graph — local-first, evidence-linked.
Basic information
Config
Add this server to your MCP-compatible client using the configuration below.
{
"mcpServers": {
"tokenknows": {
"command": "python3",
"args": [
"-m",
"venv",
".venv",
"&&",
".venv/bin/pip",
"install",
"-e",
".[dev]"
]
}
}
}Tools
No tools detected
We auto-extract tools from the README. The maintainer can list them under a ## Tools heading to populate this section.
Overview
What is TokenKnows?
TokenKnows automatically captures AI coding sessions (from tools like Claude Code, Codex, and Cursor) and distills them into structured knowledge assets such as weekly reports, ADRs, and a knowledge graph. It is for developers who want to preserve decisions and insights from AI pair-programming sessions.
How to use TokenKnows?
Install the TokenKnows backend (FastAPI) and web UI, then add the MCP plugin for your AI tool (Claude Code, Codex, Cursor, or VS Code). The plugin provides MCP tools like submit_session_events and distill_document, plus slash commands like /tokenknows:weekly. Run the backend with uvicorn app.main:app and frontend with npm run dev.
Key features of TokenKnows
- Captures AI coding sessions from Claude Code, Codex, Cursor, and more.
- Distills into seven asset types including weekly reports, ADRs, and incident reviews.
- Every paragraph traces back to original PR, conversation, or commit.
- Local-first with three-layer LLM egress gate for privacy.
- Can run fully offline with Ollama and zero cloud keys.
- All collectors run locally without webhooks or tunnels.
Use cases of TokenKnows
- Automatically generate weekly reports from AI coding sessions.
- Create Architecture Decision Records (ADRs) from design discussions.
- Produce incident reviews with evidence from PRs and commits.
- Build a knowledge graph of entities from multiple sessions.
- Reusable agent skills (SKILL.md) for future automation.
FAQ from TokenKnows
Which AI tools does TokenKnows support?
It captures sessions from Claude Code, Codex, Cursor, VS Code, GitHub (PRs/commits/issues), and local documents.
Does TokenKnows require cloud LLM keys?
No. It can run fully locally with Ollama; optional cloud providers (Anthropic, OpenAI, MiniMax) require all three egress switches to be on.
How are citations ranked in documents?
Citations are ranked by cosine × trust × recency, requiring at least two distinct sources. Each paragraph links back to the original evidence.
Is TokenKnows local-first?
Yes. Zero egress by default; a three-layer LLM egress gate (instance, project, task) controls any cloud calls, with full audit logging.
What platforms are supported?
macOS provides a full experience with auto-starting collectors. Linux runs backend, frontend, and collectors manually. Windows is untested; WSL2 is recommended.
More Memory & Knowledge MCP servers
MemoryMesh
CheMiguel23A knowledge graph server that uses the Model Context Protocol (MCP) to provide structured memory persistence for AI models.
Semantic Scholar MCP Server
YUZongminA FastMCP server implementation for the Semantic Scholar API, providing comprehensive access to academic paper data, author information, and citation networks.
mcp-local-rag
nkapila6"primitive" RAG-like web search model context protocol (MCP) server that runs locally. ✨ no APIs ✨
JupyterMCP - Jupyter Notebook Model Context Protocol Integration
jjsantos01A Model Context Protocol (MCP) for Jupyter Notebook
Context Portal MCP (ConPort)
GreatScottyMacContext Portal (ConPort): A memory bank MCP server building a project-specific knowledge graph to supercharge AI assistants. Enables powerful Retrieval Augmented Generation (RAG) for context-aware development in your IDE.
Comments