MCP 服务器

Librarian

Provides AI agents with persistent knowledge storage, enabling them to store, search, and retrieve text, documents, and files using semantic and keyword search via MCP tools.

README

Agent Library

A personal knowledge library for AI agents, built on Arcade for the Model Context Protocol (MCP).

Status. Agent Library is a community project built by Arcade.dev engineers. It is not an official Arcade.dev product and is not covered by Arcade.dev support or SLAs. See Support below for what to expect.

Naming. The package is Agent Library (agent-library on PyPI, librarian as the Python module). The MCP server inside it identifies itself as Librarian — that's the role your agent talks to, and it's why the exposed tools are named Librarian_SearchLibrary, Librarian_AddToLibrary, and so on.

Overview

Agent Library provides AI agents with persistent storage for text, documents, and knowledge. Agents can store information and retrieve it later through semantic and keyword search, maintaining context across conversations.

graph LR
    A[Agent Stores Info] --> B[Parser]
    B --> C[Chunker]
    C --> D[Embedder]
    D --> E[(SQLite + vec)]
    F[Agent Queries] --> G[Hybrid Search]
    E --> G
    G --> H[Relevant Context]

Features

Persistent knowledge storage for AI agents
SQLite storage with sqlite-vec for vector search
Full-text search using FTS5 with BM25 ranking
Hybrid search combining semantic and keyword matching
Max Marginal Relevance (MMR) for diverse results
Configurable embedding models (local or OpenAI-compatible API)
Header-aware text chunking with overlap
Time-bounded search filters
CLI and MCP server interfaces

Multi-Modal Support

Agent Library supports indexing and searching across multiple file types:

Asset Type	File Extensions	Features
Text	`.md`, `.txt`	Frontmatter extraction, header-aware chunking
Code	`.py`, `.js`, `.ts`, `.go`, `.rs`, `.java`, `.cpp`, and more	Symbol extraction (classes, functions, methods)
PDF	`.pdf`	Page-based text extraction
Image	`.png`, `.jpg`, `.jpeg`, `.gif`, `.webp`	Metadata and EXIF extraction, optional OCR

Installation

The recommended way to install Agent Library is as a uv tool, which gives you the libr CLI and the MCP server in an isolated environment:

uv tool install 'agent-library[all]'

The [all] extra pulls in optional support for PDFs, images, OCR, and code-aware embeddings. If you'd rather install only what you need:

uv tool install agent-library                # core only (text + code)
uv tool install 'agent-library[pdf]'         # add PDF support
uv tool install 'agent-library[vision]'      # add image support
uv tool install 'agent-library[ocr]'         # add OCR for image-based PDFs

Installing from source

If you want to contribute or run from a clone:

git clone https://github.com/arcadeai-labs/agent-library.git
cd agent-library
./setup.sh

Or install manually with the dev extras:

uv pip install -e ".[dev,all]"

CLI Usage

# Add files to the library
libr add ~/notes

# Search the library
libr search "machine learning concepts"

# List sources
libr list

# View library statistics
libr index

# Rebuild the index
libr index build

MCP Server

Start the server for AI assistant integration:

# stdio transport (Claude Desktop, CLI)
libr serve stdio

# HTTP transport (Cursor, VS Code)
libr serve http --port 8000

See the Arcade MCP documentation for integration details.

Available Tools

Core Tools (always enabled):

Tool	Description
`Librarian_SearchLibrary`	Unified search with mode selection (hybrid/semantic/keyword), asset type filtering, and timeframe support
`Librarian_AddToLibrary`	Store new content in the library
`Librarian_UpdateLibraryDoc`	Update existing content
`Librarian_ReadFromLibrary`	Read full document content
`Librarian_RemoveFromLibrary`	Remove content from the library
`Librarian_ListLibraryContents`	List all stored content
`Librarian_IndexDirectoryToLibrary`	Bulk import files from a directory

Optional Tools (enable with LIBRARIAN_ENABLE_OPTIONAL_TOOLS=true):

Tool	Description
`Librarian_GetLibraryOverview`	Inspect the library — `view` selects `sections` (default; storage locations + doc counts), `stats` (totals + config), or `tree` (recursive filesystem walk)
`Librarian_SuggestLibraryLocation`	AI-powered suggestions for where to store content

Configuration

Set via environment variables:

Variable	Default	Description
`DOCUMENTS_PATH`	`./documents`	Root directory for files
`DATABASE_PATH`	`~/.librarian/index.db`	SQLite database location
`EMBEDDING_PROVIDER`	`openai`	`local` or `openai`
`EMBEDDING_MODEL`	`all-MiniLM-L6-v2`	Local model name
`OPENAI_API_BASE`	`http://localhost:7171/v1`	OpenAI-compatible API URL
`OPENAI_EMBEDDING_MODEL`	`qwen3-embedding-06b`	API model name
`CHUNK_SIZE`	`512`	Max characters per chunk
`CHUNK_OVERLAP`	`50`	Overlap between chunks
`SEARCH_LIMIT`	`10`	Default results limit
`MMR_LAMBDA`	`0.7`	MMR diversity (0=diverse, 1=relevant)
`HYBRID_ALPHA`	`0.7`	Vector vs keyword weight (1=vector only)

Project Structure

librarian/
├── cli.py           # Command-line interface
├── server.py        # MCP server and tool definitions
├── config.py        # Configuration management
├── indexing.py      # Document indexing service
├── types.py         # Shared type definitions
├── storage/
│   ├── database.py  # SQLite operations
│   ├── vector_store.py  # sqlite-vec search
│   └── fts_store.py     # FTS5 search
├── processing/
│   ├── embed/       # Embedding providers
│   ├── parsers/     # Document parsers (md, code, pdf, image)
│   └── transform/   # Text chunking
├── retrieval/
│   └── search.py    # Hybrid search + MMR
└── utils/
    └── timeframe.py # Time filter utilities

Development

make install    # Install dependencies
make test       # Run tests
make lint       # Run linter
make format     # Format code
make typecheck  # Type checking
make check      # All checks
make evals      # Run evaluations

Resources

Arcade.dev - Build AI-native applications
Arcade Documentation - Integration guides and API reference

Support

Agent Library is a community project built by Arcade.dev engineers. It is not an official Arcade.dev product and is not covered by Arcade.dev support or SLAs.

GitHub issues are monitored on a best-effort basis. Expect response times in days-to-weeks, not hours.
Pull requests are welcome. See CONTRIBUTING.md for how to file one and what we're likely to accept.
Security vulnerabilities should be reported privately per SECURITY.md, not via public issues.

If you're using Agent Library in something important, snapshot the SQLite index (~/.librarian/index.db) before anything you can't easily redo. The index is a single file — cp is your friend.