graphmemory
An MCP server that builds a semantic graph memory from a project directory, indexing documentation and code into graph structures and exposing 70+ MCP tools for search, knowledge management, task management, and more.
README
graphmemory
An MCP server that builds a semantic graph memory from a project directory. Indexes markdown docs, TypeScript/JavaScript source code, and all project files into graph structures, then exposes them as 70 MCP tools + REST API + Web UI.

Quick start
npm install -g @graphmemory/server
cd /path/to/my-project
graphmemory serve
That's it. No config file needed — the current directory becomes your project. Open http://localhost:3000 for the web UI.
The embedding model (~560 MB) downloads on first startup and is cached at ~/.graph-memory/models/.
Connect an MCP client
Claude Code:
claude mcp add --transport http --scope project graph-memory http://localhost:3000/mcp/my-project
Claude Desktop — add via Settings > Connectors, enter the URL:
http://localhost:3000/mcp/my-project
Cursor / Windsurf / other clients — enter the URL directly in settings:
http://localhost:3000/mcp/my-project
The project ID is your directory name. Multiple clients can connect simultaneously.
With a config file
For multi-project setups, custom embedding models, auth, or workspaces — create graph-memory.yaml:
projects:
my-app:
projectDir: "/path/to/my-app"
docs-site:
projectDir: "/path/to/docs"
graphs:
code:
enabled: false
graphmemory serve --config graph-memory.yaml
See docs/configuration.md for full reference and graph-memory.yaml.example for all options.
Docker
docker run -d \
--name graph-memory \
-p 3000:3000 \
-v $(pwd)/graph-memory.yaml:/data/config/graph-memory.yaml:ro \
-v /path/to/my-app:/data/projects/my-app:ro \
-v graph-memory-models:/data/models \
ghcr.io/graph-memory/graphmemory-server
Docker Compose:
services:
graph-memory:
image: ghcr.io/graph-memory/graphmemory-server
ports:
- "3000:3000"
volumes:
- ./graph-memory.yaml:/data/config/graph-memory.yaml:ro
- /path/to/my-app:/data/projects/my-app
- models:/data/models
restart: unless-stopped
depends_on:
redis:
condition: service_healthy
redis:
image: redis:7-alpine
restart: unless-stopped
volumes:
- redis-data:/data
healthcheck:
test: ["CMD", "redis-cli", "ping"]
interval: 10s
timeout: 3s
retries: 3
volumes:
models:
redis-data:
Redis is optional. Remove the
redisservice anddepends_onif you don't need shared session store or embedding cache.
See docs/docker.md for details.
What it does
| Feature | Description |
|---|---|
| Docs indexing | Parses markdown into heading-based chunks with cross-file links and code block extraction |
| Code indexing | Extracts AST symbols (functions, classes, interfaces) via tree-sitter |
| File index | Indexes all project files with metadata, language detection, directory hierarchy |
| Knowledge graph | Persistent notes and facts with typed relations and cross-graph links |
| Task management | Kanban workflow with priorities, assignees, and cross-graph context |
| Skills | Reusable recipes with steps, triggers, and usage tracking |
| Hybrid search | BM25 keyword + vector cosine similarity with BFS graph expansion |
| Real-time | File watching + WebSocket push to UI |
| Multi-project | One process manages multiple projects from a single config |
| Workspaces | Share knowledge/tasks/skills across related projects |
| Auth & ACL | Password login (JWT), API keys, OAuth 2.0 (PKCE), 4-level access control |
70 MCP tools
| Group | Tools |
|---|---|
| Context | get_context |
| Docs | docs_list_files, docs_get_toc, docs_search, docs_get_node, docs_search_files, docs_find_examples, docs_search_snippets, docs_list_snippets, docs_explain_symbol, docs_cross_references |
| Code | code_list_files, code_get_file_symbols, code_search, code_get_symbol, code_search_files |
| Files | files_list, files_search, files_get_info |
| Knowledge | notes_create, notes_update, notes_delete, notes_get, notes_list, notes_search, notes_create_link, notes_delete_link, notes_list_links, notes_find_linked, notes_add_attachment, notes_remove_attachment |
| Tasks | tasks_create, tasks_update, tasks_delete, tasks_get, tasks_list, tasks_search, tasks_move, tasks_reorder, tasks_link, tasks_create_link, tasks_delete_link, tasks_find_linked, tasks_add_attachment, tasks_remove_attachment |
| Epics | epics_create, epics_update, epics_delete, epics_get, epics_list, epics_search, epics_link_task, epics_unlink_task |
| Skills | skills_create, skills_update, skills_delete, skills_get, skills_list, skills_search, skills_recall, skills_bump_usage, skills_link, skills_create_link, skills_delete_link, skills_find_linked, skills_add_attachment, skills_remove_attachment |
Web UI
Dashboard, Knowledge (notes CRUD), Tasks (kanban board with drag-drop), Skills (recipes), Docs browser, Code browser (symbols, edges, source), Files browser, Prompts (AI prompt generator), Search (cross-graph), Tools (MCP explorer), Help.
Light/dark theme. Real-time WebSocket updates. Login page when auth is configured.
Authentication
users:
alice:
name: "Alice"
email: "alice@example.com"
apiKey: "mgm-key-abc123"
passwordHash: "$scrypt$65536$..." # generated by: graphmemory users add
server:
jwtSecret: "your-secret"
defaultAccess: rw
redis: # optional: session store + embedding cache
enabled: true
url: "redis://localhost:6379"
- UI login: email + password → JWT cookies (httpOnly, SameSite=Strict)
- API access:
Authorization: Bearer <apiKey> - OAuth 2.0: Authorization Code + PKCE (S256) with frontend consent page at
/ui/auth/authorize; also supports client credentials and refresh tokens - OAuth endpoints:
/api/oauth/userinfo,/api/oauth/introspect,/api/oauth/revoke,/api/oauth/end-session - ACL: graph > project > workspace > server > defaultAccess (
deny/r/rw) - Redis (optional): when
server.redisis configured, used for OAuth session store and embedding cache
Development
npm run dev # tsc --watch (backend)
cd ui && npm run dev # Vite on :5173, proxies /api → :3000
npm test # 1809 tests across 45 suites
Documentation
Full documentation is in docs/:
- Concepts: docs indexing, code indexing, tasks, skills, knowledge, file index
- Architecture: system architecture, graphs overview, search algorithms, embeddings
- API: REST API, MCP tools guide, WebSocket
- Operations: CLI, configuration, Docker, npm
- Security: authentication, security
- UI: architecture, features, patterns
- Development: testing, API patterns
Contributing
Contributions are welcome! See CONTRIBUTING.md for guidelines.
For security vulnerabilities, see SECURITY.md.
License
Elastic License 2.0 (ELv2) — free to use, modify, and self-host. Not permitted to offer as a managed/hosted service.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.