Saga

Saga

A local-first MCP server for document ingestion and semantic search, providing tools to add, search, and retrieve documents, chunks, and code blocks.

Category
访问服务器

README

npm version

Saga MCP Server

Saga is a local-first MCP server for document ingestion and semantic search. It provides document, chunk, and code-block retrieval tools over MCP (stdio or HTTP stream).

Requirements

  • Node.js 22+
  • npm 10+

Install

From npm

npm install -g @maxinedotdev/saga

From source

git clone https://github.com/maxinedotdev/saga.git
cd saga
npm install
npm run build

Server Setup

Create ~/.saga/saga.toml:

[server]
transport = "httpStream"
base_dir = "~/.saga"

[server.http]
host = "127.0.0.1"
port = 8080
endpoint = "/mcp"
stateless = true

[env]
MCP_EMBEDDING_BASE_URL = "http://127.0.0.1:1234/v1"
MCP_EMBEDDING_MODEL = "llama-nemotron-embed-1b-v2"

Notes:

  • stateless = true is recommended and avoids session/SSE stream churn.
  • If you want stdio instead of HTTP, set transport = "stdio" and use command mode in your MCP client.

Run

Foreground

node dist/server.js --config ~/.saga/saga.toml

Background service

macOS:

npm run service:install:mac
npm run service:status:mac

Linux (systemd user service):

npm run service:install:linux
npm run service:status:linux

Windows (run terminal as Administrator):

npm run service:install:windows
npm run service:status:windows

MCP Client Config

URL mode (HTTP stream)

Use:

http://127.0.0.1:8080/mcp

Kilo (mcp_settings.json):

{
  "saga": {
    "type": "streamable-http",
    "url": "http://127.0.0.1:8080/mcp",
    "timeout": 600,
    "disabled": false
  }
}

Codex (~/.codex/config.toml):

[mcp_servers.saga]
enabled = true
url = "http://127.0.0.1:8080/mcp"

Command mode (stdio)

{
  "mcpServers": {
    "saga": {
      "command": "saga",
      "env": {
        "MCP_BASE_DIR": "~/.saga",
        "MCP_EMBEDDING_BASE_URL": "http://127.0.0.1:1234/v1",
        "MCP_EMBEDDING_MODEL": "llama-nemotron-embed-1b-v2"
      }
    }
  }
}

Tools

  • add_document
  • crawl_documentation
  • search_documents
  • get_document
  • list_documents
  • get_uploads_path
  • process_uploads
  • list_uploads_files
  • delete_document
  • delete_crawl_session
  • get_context_window
  • query
  • search_code_blocks
  • get_code_blocks
  • search_documents_with_ai (only registered when AI provider config is present)

Key Config

Config source precedence:

  1. Environment variables
  2. TOML config (MCP_CONFIG_TOML or SAGA_CONFIG_TOML)
  3. Defaults

Common variables:

  • MCP_BASE_DIR (default ~/.saga)
  • MCP_TRANSPORT (stdio or httpStream)
  • MCP_HTTP_HOST (default 127.0.0.1)
  • MCP_HTTP_PORT (default 8080)
  • MCP_HTTP_ENDPOINT (default /mcp)
  • MCP_HTTP_PUBLIC (true binds to 0.0.0.0)
  • MCP_HTTP_STATELESS (default HTTP behavior is stateless; set false for session mode)
  • MCP_EMBEDDING_BASE_URL (required)
  • MCP_EMBEDDING_MODEL
  • MCP_AI_BASE_URL (enables search_documents_with_ai)
  • MCP_AI_MODEL
  • MCP_AI_API_KEY

Development

npm run dev
npm run build
npm run test
npm run test:unit
npm run test:integration
npm run test:benchmark
npm run test:coverage
npm run db:init
npm run db:drop
npm run db:benchmark

Data Layout

By default Saga stores data under ~/.saga/:

  • uploads/
  • data/
  • lancedb/
  • logs/ (when MCP_LOG_TO_FILE=true)

Troubleshooting

If logs show lines like establishing new SSE stream for session ID ..., Saga is running HTTP stream in stateful mode. Set MCP_HTTP_STATELESS=true (or server.http.stateless = true in TOML) and restart.

Saga does not automatically publish logs to a VS Code UI panel. Use process stdout/stderr or service log files:

  • macOS/Linux service scripts: ~/.saga/logs/saga-mcp.out.log and ~/.saga/logs/saga-mcp.err.log
  • Foreground run: terminal output
  • Optional file logging: set MCP_LOG_TO_FILE=true (writes under ~/.saga/logs/)

License

MIT

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选