An open-source context retrieval layer for AI agents and RAG systems, supporting unified connection to 50+ data sources, real-time sync, multi-modal semantic search, and native MCP protocol support.
Airweave is an open-source context retrieval layer designed for AI agents and RAG systems, addressing the "context fragmentation" problem where enterprise data is scattered across multiple heterogeneous systems. A Y Combinator-backed startup that has raised $6M in seed funding (led by FCVC, with participation from LUX Capital, Y Combinator, Orange Collective, Pioneer Fund, and others).
Core Capabilities
- Unified Data Connection: Connects to 50+ apps and databases (Stripe, Notion, Google Drive, etc.) with unified authentication, ingestion, sync, and indexing
- Multi-modal Search: Semantic search, keyword search, hybrid search, time-aware search, intent understanding
- Real-time Data Sync: Continuous data source synchronization ensures agents always access up-to-date context
- Native MCP Support: Official MCP Server (stdio + Streamable HTTP) compatible with Claude Desktop, Cursor, and other MCP clients
- Multi-tenant Architecture: Redis session management, rate limiting, OAuth2 authentication
Integration Methods
- Python SDK / TypeScript SDK
- Python CLI (rich interactive terminal & JSON pipe output)
- MCP Server (local stdio mode / hosted HTTP mode)
Deployment Modes
- Cloud-hosted: app.airweave.ai
- Self-hosted: Docker Compose (development) / Kubernetes (production)
Architecture Overview Backend built on FastAPI, PostgreSQL for metadata, Vespa as the vector retrieval engine, Temporal for data sync workflow orchestration, Redis for messaging and session management. Frontend uses React + ShadCN UI. MCP Server is a standalone TypeScript/Node.js service. Integrates with LangChain, Composio, Pipedream, and other agent frameworks. Language distribution: Python (78.3%), TypeScript (16.8%), MDX (3.9%). MIT License.