Llm Wiki Mcp

flsteven87/llm-wiki-mcp

Implements Karpathy's LLM wiki pattern as an MCP server with four tools: wiki_read, wiki_write_page, wiki_log_append, and wiki_inventory. The server handles atomic writes, etag-based conflict detection, and path containment while leaving schema definition to your own CLAUDE.md file that grows with your domain. Claude Code users get four bundled skills (wiki-init, wiki-ingest, wiki-query, wiki-lint) that encode the full workflow. Other clients like Claude Desktop get the raw tools but have to derive bookkeeping steps themselves. Useful when you want persistent markdown memory that survives sessions and enforces mechanical correctness without imposing a fixed content schema. Python API lets you swap the local filesystem backend for SQLite, Notion, or test fakes by implementing the WikiStorage protocol.

tools: 4
transport: STDIO
last updated: Apr 8, 2026

View on GitHub

Tools

4 tools

Public tool metadata for what this MCP can expose to an agent.

wiki_readRead one page. Returns body, parsed frontmatter, outgoing links, etag.

Read one page. Returns body, parsed frontmatter, outgoing links, etag.

No parameter schema in public metadata yet.

wiki_write_pageAtomic create or update with etag CAS. Pass etag=null to create, the read etag to update.

Atomic create or update with etag CAS. Pass etag=null to create, the read etag to update.

No parameter schema in public metadata yet.

wiki_log_appendAppend one entry to log.md in Karpathy's ## [YYYY-MM-DD] op | Title format.

Append one entry to log.md in Karpathy's ## [YYYY-MM-DD] op | Title format.

No parameter schema in public metadata yet.

wiki_inventorySnapshot the whole graph: pages, frontmatter, link edges, log entries, plus an optional plain-text mention scan for backlink audits.

Snapshot the whole graph: pages, frontmatter, link edges, log entries, plus an optional plain-text mention scan for backlink audits.

No parameter schema in public metadata yet.

Readme

llm-wiki-mcp

Persistent markdown wiki for your AI agent, built on Karpathy's LLM wiki gist. Four MCP tools (wiki_read, wiki_write_page, wiki_log_append, wiki_inventory) plus four Claude Code skills (wiki-init, wiki-ingest, wiki-query, wiki-lint). stdio transport, local filesystem.

The server handles the boring layer LLMs keep getting wrong: atomic writes, etag conflict checks, append-only log integrity, path containment. The skills give the agent a workflow to follow. The wiki schema lives in your own wiki/CLAUDE.md and grows with your domain. There is no Layer 3 schema validation in the server.

Status: alpha (v0.1.1). Local backend only. MIT licensed.

Quick start

Requires Python 3.11+ and uv.

Pick an absolute path for the wiki folder. The server creates pages/ and log.md under it on first run if they don't exist:

uvx llm-wiki-mcp --wiki-root /absolute/path/to/wiki

Wire it into your MCP client.

Claude Code:

claude mcp add llm-wiki -- uvx llm-wiki-mcp --wiki-root /absolute/path/to/wiki

View the full README on GitHub

Details

Repository: github.com/flsteven87/llm-wiki-mcp
Categories: AI & LLM Tools, Documents & Knowledge
Package: llm-wiki-mcp
Transport: STDIO
Registry: MCP registry, active

Related AI & LLM Tools MCP servers

View all

Agent Output Guardagenson-tools/agent-output-guard-mcpValidate and verify data from other agents before acting on it. Zero LLM costs.0stars

Vector Search APIbr0ski777/vector-search-x402In-memory vector search with TF-IDF and cosine similarity. x402 micropayment.0stars

Agent Toolbelt Mcpmarras0914/agent-toolbelt19 focused API tools for AI agents: contract review, prompt optimizer, context packing, and more.0stars

Agent Costcentermdfifty50-boop/agent-costcenter-mcpMulti-agent billing and cost allocation tracking0stars

Traced Llm Proxymikerawsonnz/gosce-agentsProxy Gemini (Vertex AI) completions wrapped in OpenTelemetry trace spans; returns the answer plus t0stars

agent-godmodemdvohra/agent-godmodeSandboxed workspace MCP: file ops, argv-only run_command, list_files; BYO LLM prompts.0stars