# Airweave Search

**Category:** 🔎 Search & Data Extraction  
**Repository:** https://github.com/airweave-ai/airweave  
**GitHub Stars:** 6544  
**Views:** 2  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/airweave-search

## Description
MCP server for searching Airweave collections with natural language queries.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `uvx` (confidence: high):

```json
"mcpServers": {
  "airweave-search": {
    "command": "uvx",
    "args": ["airweave-sdk"]
  }
}
```

## Documentation

## What Airweave Search MCP server does

The Airweave Search MCP server exposes Airweave’s unified retrieval layer to MCP-compatible AI agents. It is designed for searching collections that contain data synchronized from connected applications, databases, and documents. Instead of requiring an agent to query each source independently, Airweave provides one collection-oriented search surface.

Airweave supports more than 50 integrations, including services such as GitHub, GitLab, Gmail, Google Drive, Jira, Linear, Notion, Slack, Salesforce, SharePoint, and Stripe. The available sources depend on the integrations configured in Airweave. The MCP server searches the resulting Airweave collections; it is not described as a direct API client for each individual source.

## How it works

Airweave sits between source systems and the agent. First, applications, databases, or documents are connected to Airweave. Airweave then handles synchronization and indexing, making the collected data available through its retrieval layer. An agent sends a natural-language request through MCP, and the Airweave Search MCP server searches the relevant collection for matching context.

This arrangement is useful when several systems contribute to the information an agent needs. It also centralizes retrieval rather than requiring a separate search implementation for every connected application. The repository describes access through MCP alongside SDKs, a REST API, and native integrations with agent frameworks.

## Setup and configuration

Airweave can be used as a hosted service at `app.airweave.ai` or run locally from the repository. The documented self-hosted path requires Docker and Docker Compose:

```bash
git clone https://github.com/airweave-ai/airweave.git
cd airweave
./start.sh
```

The local application is available at `http://localhost:8080` after startup. The startup process creates an environment file from the example configuration, generates required secrets, starts the services with health checks, and can prompt for OpenAI or Mistral API keys. Initial startup may take several minutes while services become healthy.

The repository also provides startup options to restart services, skip the frontend for a backend-only deployment, or destroy the local setup. Docker must be running before these commands are used.

## Tools and capabilities

The documented capability is natural-language search over Airweave collections through MCP. This makes the Airweave Search MCP server suitable for agents that need relevant context from synchronized business and development data.

## Limitations and notes

Search quality and available results depend on which sources have been connected, synchronized, and indexed in Airweave. The MCP server does not eliminate the need to configure those sources.

_Full upstream README: https://allmcps.com/mcp/airweave-search/readme_