Integrations

DeepFellow integrates with leading MCPs and external tools to enable secure use of market-standard models within your private context. This integration ensures the secure use of models while maintaining the privacy of your internal data.

Secure Integration with MCPs

DeepFellow provides a secure way to integrate with MCPs through custom connectors, which ensure compatibility with your tech stack without exposing internal data. This approach allows you to leverage market-standard models while maintaining control over your internal data.

API for MCP Connections

DeepFellow offers an API to connect MCPs in two ways:

• Using DeepFellow's OpenAI-compatible /v1 group, which covers chat completions, embeddings, images, audio, fine-tuning, vector stores, and /v1/responses for tool and toolbox calling. See the Access Matrix for the credential requirements and OpenAI API via DeepFellow for usage examples.
• Creating your own toolbox.

Toolbox is a set of tools accessible from outside of your infrastructure using Model Context Protocol (MCP). You can create multiple toolboxes using the following tools:

• File Search (file_search)
• Image Generation (image_generation)
• External MCP (mcp)
• Infra MCP (infra-mcp)
• Websearch (websearch)
• Custom (custom-mcp)

Every toolbox exposes its tools over standard MCP transports, Streamable HTTP and SSE, protected by a bearer token. See Supported MCP Protocols and Tool Calling for the endpoint URLs and authentication details.

Internal Data Safety

By using DeepFellow's secure integration approach, you can ensure the safety of your internal data. This approach allows you to use market-standard models while maintaining control over your internal data, ensuring that your data remains private and secure.