SkillFoxx

LiteGraph MCP

A .NET graph database with an MCP server of 185 tools: Claude creates nodes and edges, searches vectors and traverses the graph.

  • 133 ★ on the repository
  • Last commit
MCP server

Extended access

Why this level

  • The agent can change and delete graph data, and admin tools back up and flush data.
  • The default admin token is public in the documentation.
All reasons and checks

litegraphdb/litegraphAuthor:litegraphdb

Install

Manual install

git clone https://github.com/litegraphdb/litegraph.git

Then `dotnet build` in `src`, `./LiteGraph.McpServer install` and starting the servers; details are in `docs/CLAUDE_MCP.md`.

Access and checks

How we review

Extended accessThe tool runs code or makes network calls and only reads external data.

  • The agent can change and delete graph data, and admin tools back up and flush data.
  • The default admin token is public in the documentation.

This is third-party code. Review the repository files before installing.

What it does

LiteGraph is a graph database for .NET with relational data, tags, labels, JSON and vector search. Its built-in MCP server gives an agent more than 185 tools: managing tenants, graphs, nodes and edges, vector search, algorithms such as PageRank and community detection, and JSONL export and import. It listens on HTTP at `http://localhost:8702/mcp`, plus TCP and WebSocket. The server forwards calls to the LiteGraph REST server and caches nothing. Storage is a choice: SQLite for local work or PostgreSQL with pgvector for production. A dashboard and a chat with the graph come alongside.

Who it is for. Developers on .NET and analysts who build knowledge graphs and want to manage them in conversation with an agent.

Good fit when

  • An agent builds a graph of links between people, documents or services from a description.
  • Semantic search over nodes with embeddings is needed next to graph queries.
  • A graph is exported to JSONL or GEXF for transfer and analysis.
  • Data of several teams is separated by tenants with roles.

Not a fit when

  • A ready package without building from source is required.
  • A plain relational table is enough and no traversal is needed.
  • An agent must not be trusted with backup, restore and data flush.

Example request

Create a tenant demo-tenant and a graph social-network, add five nodes and link them pairwise, then show the graph summary.

MCP

Transport
http
Authentication
API key

Environment variables

LITEGRAPH_ENDPOINT required
Address of the LiteGraph REST server that the MCP server forwards calls to.
LITEGRAPH_API_KEY required, secret
API key (bearer token) for the REST server.

Limitations

There is no ready package: the MCP server is built from source (.NET SDK 8 or newer) and run next to the LiteGraph REST server. The repository states no license. The `admin_*` tools back up, restore and flush data, so give the agent a key with minimal rights. The call rate limit is off by default (`ToolCallsPerSecond` is 0). Production deployments on PostgreSQL need the pgvector extension. The MCP server stores no data and works only while the REST server it calls is running.

How to disable. Stop the MCP server and remove the connection with `claude mcp remove` or by editing the client configuration. Data stays in the database and is removed separately.

FAQ

How do I connect Claude Code?

Run `./LiteGraph.McpServer install` from the build folder: it registers the server with Claude Code and creates a `litegraph` agent definition. The `--dry-run` flag previews the changes.

Why is the `graph/get` tool not found?

Since version 10.1 names use underscores: `graph_get`. Replace each slash with an underscore; arguments and results are unchanged.

Which addresses and ports does the server use?

HTTP MCP runs at `http://localhost:8702/mcp`, TCP on port 8703 and WebSocket at `ws://localhost:8704/mcp`. Change them in `litegraph-mcp.json` or with `MCP_HTTP_*` variables.

Which storage backends are supported?

SQLite for local work with the in-process HnswLite vector index, and PostgreSQL with pgvector for production, including a multi-node cluster.

How do I run the server in Docker?

The `docker/` folder holds three Docker Compose setups: a single node on SQLite, a single node on PostgreSQL and a three-node cluster. Each includes the MCP server and the dashboard.

Discussion

No questions or reviews yet.

MCP servers

KGLite keeps a knowledge graph in one file, and its bundled MCP server gives agents Cypher queries with no separate database.

MCP serverExtended access77

A knowledge graph with an MCP server: the agent searches concepts by meaning and sees how well each is supported.

MCP serverExtended access131
Official

The official Graphlit MCP server: an agent collects material from Slack, GitHub, mail and websites into a knowledge base and searches it

MCP serverBroad access379

GraphJin MCP

GraphJin

Official

GraphJin's MCP server gives an agent safe, schema-aware access to SQL and MongoDB instead of raw database credentials

MCP serverBroad access3.2K

MCP server for the Flexible GraphRAG platform: ingest from 14 data sources, hybrid search and AI answers over a knowledge graph

MCP serverBroad access200
Official

A CLI and MCP server that extracts structured knowledge from documents: graphs, hypergraphs, and fact search with sources

CLIExtended access4.2K

README in short

The repository holds the database core, REST and MCP servers, the `lg` console, a Next.js dashboard, SDKs for C#, Python and JavaScript and three Docker Compose setups.

README badge

Report a problem