Swarms MCP Documentation Server

🐝 Swarms MCP Documentation Server

Version 2.2

📖 Description

This program is an Agent Framework Documentation MCP Server built on FastMCP, designed to enable AI agents to efficiently retrieve information from your documentation database. It combines hybrid semantic (vector) and keyword (BM25) search, chunked indexing, and a robust FastMCP tools API for seamless agent integration.

Key Capabilities:

Efficient, chunk-level retrieval using both semantic and keyword search
Agents can query, list, and retrieve documentation using FastMCP tools
Local-first, low-latency design (all data indexed and queried locally)
Automatic reindexing on file changes
Modular: add any repos to corpora/, support for all major filetypes
Extensible: add new tools, retrievers, or corpora as needed

Main modules:

embed_documents.py → Loads, chunks, and embeds documents
swarms_server.py → Brings up the MCP server and FastMCP tools

🌟 Key Features

Hybrid Retriever 🔍: Combines semantic and keyword search.
Dynamic Markdown Handling 📄: Smart loader based on file size.
Specialized Loaders ⚙️: .py, .ipynb, .md, .txt, .yaml, .yml.
Chunk and File Summaries 📈: Displays chunk counts along with file counts.
Live Watchdog 🔥: Instantly responds to any changes in corpora/.
User Confirmation for Costs ✅: Confirms before expensive embeddings.
Healthcheck Endpoint 🚑: Ensure server is ready for use.
Local-First 🗂️: All repos indexed locally without external dependencies.
Safe Deletion Helper 🔥: Auto-delete broken/mismatched indexes.

🏗️ Version History

Version	Date	Highlights
2.2	2025‑04‑25	Split embed/load from server; full chunk counting in loading summaries
1.0	2025‑04‑25	Dynamic Markdown loader, color logs, Healthcheck tool
0.7	2025‑04‑25	Specialized file loaders for `.py`, `.ipynb`, `.md`
0.5	2025‑04‑10	OpenAI large model embeddings, extended MCP tools
0.1	2025‑04‑10	Initial version with generic loaders

📚 Managing Your Corpora (Local Repos)

Because Swarms and other frameworks are very large, full corpora are not pushed to GitHub.

Instead, you clone them manually under corpora/:

# Inside your project folder:
cd corpora/

# Clone useful frameworks:
git clone https://github.com/SwarmsAI/Swarms
git clone https://github.com/SwarmsAI/Swarms-Examples
git clone https://github.com/microsoft/autogen
git clone https://github.com/langchain-ai/langgraph
git clone https://github.com/openai/openai-agent-sdk

✅ Notes:

Add any repo — public, private, custom.
Build your own custom AI knowledge base locally.
Large repos (>500MB) are fine; all indexing is local.

🚀 Quick Start

# 1. Activate virtual environment
venv\Scripts\Activate.ps1

# 2. Install all dependencies
pip install -r requirements.txt

# 3. Configure OpenAI API Key
echo OPENAI_API_KEY=sk-... > .env

# 4. (Load and embed documents
python embed_documents.py

# 5. Start MCP server
python swarms_server.py
# If no index is found, the server will prompt you to embed documents automatically.

⚙️ Configuration

Corpus: Drop repos inside corpora/
Environment Variables:
- .env must contain OPENAI_API_KEY
Index File Support:
- Both chroma-collections.parquet and chroma.sqlite3 are supported. .parquet is preferred if both exist.
Auto-Embedding:
- If no index is found, the server will prompt you to embed and index your documents automatically.
Optional:
- Disable Chroma compaction if you prefer:
  setx CHROMA_COMPACTION_SERVICE__COMPACTOR__DISABLED_COLLECTIONS "swarms_docs"
Command-Line Flags:
- --reindex → trigger a refresh reindex during server run.

🔄 File Watching & Auto Reindexing

The MCP Server watches corpora/ for any file changes:

Any modification, creation, or deletion triggers a live reindex.
No need to restart the server.

🛠️ Available FastMCP Tools

Tool	Description
`swarm_docs.search`	Search relevant documentation chunks
`swarm_docs.list_files`	List all indexed files
`swarm_docs.get_chunk`	Get a specific chunk by path and index
`swarm_docs.reindex`	Force reindex (full or incremental)
`swarm_docs.healthcheck`	Check MCP Server status

❓ Troubleshooting

Q: I get 'No valid existing index found' when starting the server.
- A: The server will now prompt you to embed and index documents. Accept the prompt to proceed, or run python embed_documents.py manually first.
Q: Which index file is used?
- A: The server will use chroma-collections.parquet if available, otherwise chroma.sqlite3.
Q: I want to force a reindex.
- A: Run python swarms_server.py --reindex or use the swarm_docs.reindex tool.

📋 Example Usage

# Search the documentation
result = swarm_docs.search("How do I load a notebook?")
print(result)

# List all available files
files = swarm_docs.list_files()
print(files)

# Get a specific document chunk
chunk = swarm_docs.get_chunk(path="examples/agent.py", chunk_idx=2)
print(chunk["content"])

🧰 Extending & Rebuilding

Add new docs → drop into corpora/, then:
python swarms_server.py --reindex
Schema changes → (e.g. different metadata structure):
python swarms_server.py --reindex --full
Add new repo → Drop folder under corpora/, reindex.
Recommended for mostly read-only repos:
setx CHROMA_COMPACTION_SERVICE__COMPACTOR__DISABLED_COLLECTIONS "swarms_docs"

🔗 IDE Integration

Plug directly into Windsurf Cascade:

"swarms": {
  "command": "C:/…/Swarms/venv/Scripts/python.exe",
  "args": ["swarms_server.py"]
}

Then you can access swarm_docs.* tools from Cascade automations.

📦 Requirements

💡 Python 3.11 Environment Required

Create your environment explicitly:

python3.11 -m venv venv

Then install with:

pip install -r requirements.txt

✅ MCP Server Ready

After boot:

Proper loading summaries
Safe confirmation before expensive actions
Auto file watching and reindexing
Windsurf plug-in ready
Full tool coverage

You're good to cascade it! 🏄‍♂️

📈 Flow Diagram

                          +------------------+
                          |    🖥️ MCP Server  |
                          +------------------+
                                  |
     +---------------------------------------------------+
     |                                                   |
+-------------+                                     +-----------------+
|  📁 Corpora |                                     | 🔎 FastMCP Tools |
|  Folder     |                                     | (search, list,   |
|  (markdown, |                                     | get_chunk, etc.) |
|  code, etc) |                                     +-----------------+
+-------------+                                               |
      |                                                       |
+-----------------+                                   +----------------+
|  📚 Loaders      |                                   | 🧠 Ensemble    |
| (Python, MD, TXT)|                                   | Retriever (BM25|
|  Split into Chunks|                                  | + Chroma)      |
+-----------------+                                   +----------------+
      |                                                       |
+-----------------+                                   +----------------+
| ✂️ Text Splitter |                                   | 🧩 Similarity   |
| (RecursiveCharacter) |                              | Search (chunks) |
+-----------------+                                   +----------------+
      |                                                       |
+-----------------+                                   +----------------+
| 💾 Embed chunks  |  —OpenAI Embedding (small)—>    | 🛢️ Chroma Vector |
| via OpenAI API  |                                   | DB (Local Store) |
+-----------------+                                   +----------------+
      |                                                       |
+-----------------+                                   +----------------+
| 📡 Reindex Watcher|                                  | 👀 File Watchdog |
| (Auto detect      |                                  | (Auto reindex   |
| new/modified files|                                  | on file events) |
+-----------------+                                   +----------------+

This server cannot be installed

security - not tested

license - not found

quality - not tested

local-only server

The server can only run on the client's local machine because it depends on local resources.

An Agent Framework Documentation server that enables AI agents to efficiently retrieve information from documentation databases using hybrid semantic and keyword search for seamless agent integration.

Related MCP Servers

mcp-ragdocs
hannesrudolph
A
security
A
license
A
quality
An MCP server implementation that provides tools for retrieving and processing documentation through vector search, enabling AI assistants to augment their responses with relevant documentation context
Last updated -
7
62
81
TypeScript
MIT License
MCP Expert Server
crazyrabbitLTC
-
security
F
license
-
quality
A Model Context Protocol server utilizing Claude AI for generating intelligent queries and offering documentation assistance based on API documentation analysis.
Last updated -
3
2
TypeScript
RAG Documentation MCP Server
rahulretnan
-
security
F
license
-
quality
Enables AI assistants to enhance their responses with relevant documentation through a semantic vector search, offering tools for managing and processing documentation efficiently.
Last updated -
62
13
TypeScript
MCP Documentation Server
mahawi1992
-
security
F
license
-
quality
A smart documentation server that provides AI-assisted code improvement and documentation management through Claude Desktop integration.
Last updated -
7
TypeScript

View all related MCP servers

Swarms MCP Documentation Server