# Flexible GraphRAG — Backend + MCP for Docs

> An open-source GraphRAG/RAG platform with FastAPI backend + MCP server, 13 data sources, and hybrid search; verified 125★, pushed 2026-05-14.

## Install

Merge the JSON below into your `.mcp.json`:

## Quick Use

```bash
uv venv venv-3.13 --python 3.13
source venv-3.13/bin/activate
uv pip install flexible-graphrag
flexible-graphrag  # backend on http://localhost:8000 (README)
# In another terminal (README MCP quickstart):
uv venv venv-mcp --python 3.13 && source venv-mcp/bin/activate
uv pip install flexible-graphrag-mcp
flexible-graphrag-mcp --http --port 3001
```

## Intro

An open-source GraphRAG/RAG platform with FastAPI backend + MCP server, 13 data sources, and hybrid search; verified 125★, pushed 2026-05-14.

**Best for:** Teams building doc intelligence with a REST backend and MCP tools (ingest/search/query)

**Works with:** Python 3.13 + uv; backend on http://localhost:8000 and MCP server in HTTP mode on port 3001 (README)

**Setup time:** 30-65 minutes

### Key facts (verified)

- GitHub: 125 stars · 27 forks · pushed 2026-05-14.
- License: Apache-2.0 · owner avatar + repo URL verified via GitHub API.
- README-backed entrypoint: `uv pip install flexible-graphrag-mcp && flexible-graphrag-mcp --http --port 3001`.

## Main

- Treat it as two layers: backend REST API first, then MCP server connecting to it (README requires backend on :8000).

- Start with the smallest footprint: install from PyPI and run `flexible-graphrag` before enabling optional databases/extras.

- Use MCP Inspector for debugging: README suggests `npx @modelcontextprotocol/inspector` to validate tools over Streamable HTTP.

- Scale ingestion sources gradually: README states 13 data sources and 9 MCP tools; add sources only after your baseline pipeline works.

### Source-backed notes

- README states the backend is a FastAPI service at `http://localhost:8000` and includes Angular/React/Vue frontends.
- README says the MCP server provides 9 specialized tools and supports all 13 data sources via backend REST APIs.
- README MCP quickstart runs `flexible-graphrag-mcp --http --port 3001` and tests with `npx @modelcontextprotocol/inspector`.

### FAQ

- **Do I need the backend running for MCP?**: Yes — README says the MCP server connects to the backend REST API; it must be up on `http://localhost:8000`.
- **What’s a good first validation step?**: Use MCP Inspector (`npx @modelcontextprotocol/inspector`) in HTTP mode as shown in README.
- **How many tools and sources are covered?**: README states 9 MCP tools and 13 data sources supported by the pipeline.

## Source & Thanks

> Source: https://github.com/stevereiner/flexible-graphrag
> License: Apache-2.0
> GitHub stars: 125 · forks: 27

---

<!-- ZH -->

## Quick Use

```bash
uv venv venv-3.13 --python 3.13
source venv-3.13/bin/activate
uv pip install flexible-graphrag
flexible-graphrag  # backend on http://localhost:8000 (README)
# In another terminal (README MCP quickstart):
uv venv venv-mcp --python 3.13 && source venv-mcp/bin/activate
uv pip install flexible-graphrag-mcp
flexible-graphrag-mcp --http --port 3001
```

## Intro

Flexible GraphRAG 是开源 GraphRAG/RAG 平台：FastAPI 后端 + MCP server，支持 13 种数据源与混合检索/知识图谱；已验证 125★，更新于 2026-05-14。

**Best for:** 想要 REST 后端 + MCP 工具（导入/检索/问答）来做文档智能的团队

**Works with:** Python 3.13 + uv；后端 `http://localhost:8000`，MCP HTTP 模式示例端口 3001（README）

**Setup time:** 30-65 minutes

### Key facts (verified)

- GitHub：125 stars · 27 forks；最近更新 2026-05-14。
- 许可证：Apache-2.0；作者头像与仓库链接均已通过 GitHub API 复核。
- README 中可对照的入口命令：`uv pip install flexible-graphrag-mcp && flexible-graphrag-mcp --http --port 3001`。

## Main

- 把它当两层：先跑后端 REST API，再用 MCP server 连接后端（README 要求后端在 :8000）。

- 先跑通最小闭环：从 PyPI 安装并运行 `flexible-graphrag`，再按需开启额外数据库与 extras。

- 用 MCP Inspector 调试：README 推荐 `npx @modelcontextprotocol/inspector` 检查 HTTP 工具是否正常。

- 逐步扩大数据源：README 写明 13 种数据源与 9 个 MCP 工具，先跑通再加源，避免一次性堆配置。

### Source-backed notes

- README 写明后端是 FastAPI 服务，地址为 `http://localhost:8000`，并包含 Angular/React/Vue 前端。
- README 表示 MCP server 提供 9 个工具，并通过后端 API 覆盖 13 种数据源。
- README 的 MCP Quickstart 使用 `flexible-graphrag-mcp --http --port 3001`，并用 `npx @modelcontextprotocol/inspector` 测试。

### FAQ

- **MCP 需要后端一直运行吗？**：需要；README 说明 MCP server 通过后端 REST API 工作，需先启动 `http://localhost:8000`。
- **第一步怎么验证？**：按 README 用 MCP Inspector（`npx @modelcontextprotocol/inspector`）在 HTTP 模式连通测试。
- **支持多少工具和数据源？**：README 标注 MCP 9 个工具，管道支持 13 种数据源。

## Source & Thanks

> Source: https://github.com/stevereiner/flexible-graphrag
> License: Apache-2.0
> GitHub stars: 125 · forks: 27


---
Source: https://tokrepo.com/en/workflows/flexible-graphrag-backend-mcp-for-docs
Author: MCP Hub