Skip to main content
The MCP server doesn’t mirror REST endpoints one-to-one. Instead it exposes one tool per Fuse feature area, and each tool is an agent: it takes a natural-language request, carries out the work inside that feature (often several operations), and answers in plain text. That changes how you use it. Rather than orchestrating calls, describe the outcome you want and let the tool do the orchestration:
“In prospect_lists: find my list of website visitors from July and tell me how many rows it has.”
Requests should be self-contained — each call stands alone, so include the relevant names, dates, or ids rather than referring to “the list from before”. Your assistant generally does this for you, but it explains why very terse follow-ups sometimes make the tool ask for clarification.

​
The tools

Tool boundaries follow the app’s navigation. When a job spans areas — say, “campaign the companies that visited my site” — the assistant chains tools: prospect_lists to resolve the All Website Visitors list, then campaigns to attach it.

​
Deep Research

Two additional tools run Fuse’s Deep Research agent, which is asynchronous because a run takes minutes, not seconds:
  • deep_research_start begins a run and returns a runId immediately.
  • deep_research_status reports progress for a runId and returns the report when the run completes.
A typical run takes 5–20 minutes; your assistant polls deep_research_status in the meantime. Any prospect list the run produces also appears in your lists in the Fuse app, and the report is available there too.

​
Behavior and safety

  • Every tool runs with your permissions — it sees and touches only what your Fuse user can.
  • Campaigns created or edited over MCP stay drafts until you explicitly ask to launch them; an already-active campaign keeps running while you edit.
  • analytics, website_intent, and power_dialer are read-only by design.
  • Tool errors come back as short plain-text messages (for example, “The tool could not complete the request.”) rather than internal error details.

​
Rate limits

MCP usage is limited per user, independently of your REST API rate limits: Beyond a limit the server returns 429 with the code RATE_LIMIT_EXCEEDED, and your assistant should back off and retry. One conversational ask usually costs a handful of requests (a tool call plus the protocol handshake), so the limits are generous for interactive use.