Connecting Your Data

Connect your data sources to your agents through MCP servers. Support websites, GitHub repositories, documentation, knowledge bases, PDFs, and more. Your agents can search and access this content in real-time.

How it works

Set up your data source - Configure your chosen provider and connect your data sources
Get your MCP server URL - Obtain the MCP server endpoint from your provider
Register the MCP server - Register it as a tool in the Visual Builder or TypeScript SDK
Use in your agents - Your agents can now search and access your connected data

Choose a data provider

Select a provider that supports your data sources:

Inkeep

Connect 25+ data sources including websites, GitHub, Notion, Slack, Confluence, and more. Comprehensive knowledge base solution.

Pinecone

Connect DOCX, JSON, Markdown, PDF, and Text files. Ideal for document-heavy workflows and semantic search applications.

Context7

Connect GitHub, GitLab, BitBucket repositories, OpenAPI specs, and websites. Great for code documentation.

Ref

Connect GitHub repositories, PDFs, and Markdown files. Simple and focused solution for documentation.

Firecrawl

Connect websites. Great for extracting LLM-friendly content from the web.

Other options

Consider these additional options for connecting your data:

Reducto - Self-serve document processing with a free tier. Upload documents and access them via APIs or SDKs. Wrap their APIs in your own MCP server or invoke them directly using function tools.
Unstructured - Document processing platform with a free tier. Upload documents and access them via APIs or SDKs. Wrap their APIs in your own MCP server or invoke them directly using function tools.
Milvus - Open-source vector database with self-hosting options. See their MCP integration guide for setup instructions. Must manage deployment and hosting for MCP server.
SingleStore - Relational database with a managed MCP server that converts natural language queries into SQL. Learn more in their MCP server documentation.