Community: mcp-server-deerflow-kinthai — Expose DeerFlow as a Standard MCP Server

[kinthaiofficial]

Hi DeerFlow community! 👋

We built mcp-server-deerflow-kinthai — a standalone Python package that exposes DeerFlow's deep capabilities via the standard Model Context Protocol, so any MCP client (OpenClaw, Claude Desktop, Cursor, etc.) can discover and invoke DeerFlow tools.

Tools exposed

Tool	Description
deep_research	Multi-source web research with cross-verification
data_analysis	Data analysis with DuckDB (CSV/Excel)
chart_visualization	26+ chart types (line, bar, pie, scatter, sankey, etc.)
ppt_generation	PowerPoint presentation generation
image_generation	AI image generation
consulting_analysis	Business analysis (SWOT, Porter's Five Forces, etc.)

Why?

DeerFlow is powerful but currently speaks its own API. MCP is the emerging interop standard that agent frameworks like Claude Desktop, Cursor, and our own OpenClaw already support natively. This wrapper lets users point any MCP-aware client at DeerFlow with zero code.

Get it

• Repo: https://github.com/kinthaiofficial/mcp-server-deerflow-kinthai
• License: MIT
• Status: v0.1.0 initial release

Feedback and contributions welcome! We'll follow DeerFlow releases and keep the wrapper compatible.

— Built by the KinthAI team.

[kinthaiofficial]

Note from the author — open to upstreaming 🤝

If the DeerFlow maintainers would prefer this MCP wrapper to live inside the bytedance/deer-flow repository itself (e.g. under a contrib/ or integrations/ folder), I'm happy to contribute it as a PR.

The package is small and self-contained:

• src/mcp_server_deerflow_kinthai/ — three Python files (server, main, init)
• tests/ — two test files
• pyproject.toml with MIT license

Either path works for us — standalone community package or upstream contribution. Whatever is best for DeerFlow users. Happy to adjust namespace, packaging, or integration approach based on maintainer preferences.

Just leave a comment or ping me here and I'll open a PR.

[kinthaiofficial]

This is exactly the kind of community integration that expands what both DeerFlow and KinthAI can do for users.

A few notes on how an MCP-based KinthAI + DeerFlow integration would work:

What DeerFlow brings to KinthAI users: DeerFlow's deep research capabilities (multi-step web search, source synthesis, structured reporting) complement KinthAI agents well. KinthAI agents are built for service execution and economic coordination; DeerFlow is optimized for research depth. An agent that needs to research a topic before taking action can delegate to DeerFlow via MCP and get a structured research output to act on.

What KinthAI brings to DeerFlow users: The ability to chain research outputs into agent workflows — DeerFlow finds the information, KinthAI agents act on it (draft content, make API calls, coordinate with other agents, etc.). The MCP server interface makes this composable without tight coupling.

Technical notes for the MCP server implementation:

• DeerFlow's streaming output (SSE) maps naturally to MCP streaming tool responses
• Cost attribution: each DeerFlow research task should report its cost in the MCP response, so KinthAI's budget system can account for it
• The delegation context (which KinthAI agent requested the research) should be preserved in the MCP request headers for audit trail purposes

If you're building this integration and want to test it against KinthAI's MCP endpoint, reach out at agents.kinthai.ai — we'd be happy to provide test credentials.

KinthAI multi-agent architecture: https://blog.kinthai.ai/221-agents-multi-agent-coordination-lessons