Community MCP server for FalkorDB — connects LLM agents via Cypher

[Bazza-X]

Hi FalkorDB team,

I built a Python-based MCP server for FalkorDB and published it to PyPI. It's an alternative to the official FalkorDB-MCPServer (TypeScript/npm) for users in the Python/uvx ecosystem.

PyPI: https://pypi.org/project/mcp-server-falkordb/
GitHub: https://github.com/rbarrielabrystech/mcp-server-falkordb

Why a Python alternative?

The official MCP server is TypeScript/npm. Many MCP users (especially those working with Python data stacks, FastAPI backends, or uv-managed environments) prefer Python servers that install via uvx or pip without needing Node.js.

What it adds beyond the official server

• Python ecosystem: uvx mcp-server-falkordb or pip install mcp-server-falkordb — no Node.js required
• Pydantic v2 strict validation with extra="forbid" on all inputs
• Batched schema introspection via asyncio.gather (parallel label/rel count queries)
• Configurable query timeout (FALKORDB_QUERY_TIMEOUT_MS, default 30s)
• Non-fatal startup probe (warns if FalkorDB is unreachable, doesn't crash)
• Defense-in-depth read-only validator (handles backtick-quoted identifiers, block comments, line comments — on top of FalkorDB's server-side ro_query)

Tools

Tool	Purpose
graph_list	Enumerate all graphs
graph_describe	Full schema: labels, rel types, property keys, counts
graph_query	Read-only Cypher (client + server enforcement)
graph_mutate	Write Cypher with parameterized query support
graph_explore	One-call schema + sample nodes + sample edges
graph_delete	Drop a graph (requires confirm: true)

Quick start

uvx mcp-server-falkordb

Python 3.12+, MIT licensed, 98 tests, mypy strict + ruff clean.

Happy to take feedback — especially on Cypher edge cases in the read-only validator. Built with falkordb-py.

[gkorland]

Thanks for sharing!