CatchIntent
MCP server Native protocol, not a chatbot wrapper

The intent platform your AI agent can operate

CatchIntent is MCP-native. Connect Claude, Cursor, ChatGPT, or any MCP client, then run your pipeline through the AI you already use.

  • OAuth-scoped
  • Included on every plan
  • Revoke anytime
Install in
MCP-native
claude mcp add catchintent --transport http https://engine.catchintent.com/mcp --scope user
OAuth 2.1· Scoped per workspace· Revocable anytime
Why MCP-native

Built for AI agents, not chat plugins

Most "AI integrations" wrap a chatbot around an API. MCP is the open standard your agent already speaks, so your data is a first-class tool surface, not a screen it has to scrape.

Native, not a wrapper

Your AI calls tools, not URLs

MCP is a real protocol. Your agent sees typed inputs, structured outputs, and clear scopes — not a black-box plugin that scrapes the dashboard.

28+ tools · typed schemas
OAuth-scoped

Authorized, scoped, revocable

One-time OAuth 2.1 flow with workspace-scoped tokens. Read, write, and CRM scopes are granted separately and revocable from Settings any time.

9 scopes · per-workspace tokens
Any MCP client

Works wherever your AI lives

Claude Code, Codex CLI, Cursor, Gemini, Windsurf, Raycast — anything that speaks MCP. No SDK to install, no API wrapper to maintain.

HTTP transport · spec-compliant
Tool catalog

Everything your operator does, as tools

28 typed tools across six surfaces. Each one is namespaced and scoped, so your AI always knows which surface it is operating on and which scope authorises it.

3 tools

Discover leads

Search and filter the lead pool by warmth, source, agent, status, or text — the same surface your operator uses in the dashboard.

lead_search lead_get lead_update_status
6 tools

Run agents

Create, list, and tune agents from the prompt. Update ICP, targeting, signal mix, change status, and inspect every recent run — including who got delivered and why.

agent_list agent_get agent_create agent_update agent_set_status agent_runs
2 tools

Enrich + draft

Hydrate a LinkedIn lead with full profile data, then draft a personalised opener grounded in their signal — funding round, job change, hiring spike, or topic.

lead_enrich lead_draft_outreach
9 tools

Manage outreach

Enroll into an outreach sequence, reassign owners, pause for re-evaluation, resume after a reply, mark replied, or exit. Every state your unified inbox tracks.

engagement_enroll engagement_reassign engagement_pause engagement_resume engagement_mark_replied engagement_exit inbox_list
7 tools

Workspace + brands

List, create, and update products (brands). Pull the canonical ICP option lists so the AI never invents a label. Read workspace usage against plan limits.

product_list product_get product_create product_update workspace_usage workspace_icp_options
1 tools

Push to CRM

Send a scored lead to HubSpot, Close, Pipedrive, or Zoho — fanned out across whichever destinations the workspace has connected.

lead_push_to_crm

* Full schema available via MCP tool listing once authorized

Install

Wired up in 30 seconds

One server URL. One OAuth flow. Same install command shape across every major MCP client.

Install in
MCP-native
claude mcp add catchintent --transport http https://engine.catchintent.com/mcp --scope user
OAuth 2.1· Scoped per workspace· Revocable anytime
  1. Step 01

    Add the server

    One command in your MCP client of choice — or paste the server URL into a config file.

  2. Step 02

    Authorize the workspace

    Your browser opens to the dashboard. Pick the workspace, review the scopes, click Authorize. Tokens refresh automatically.

  3. Step 03

    Prompt your way through

    "Show me hot LinkedIn leads from job-change signals and draft connect notes for the top 3." Done in one message.

Compatible with every major MCP client
Claude Code Claude Desktop Codex CLI Cursor Gemini CLI ChatGPT Windsurf Copilot Raycast Claude Code Claude Desktop Codex CLI Cursor Gemini CLI ChatGPT Windsurf Copilot Raycast
Use cases

Run your pipeline by talking to it

Four prompts an operator actually uses. Every command below resolves to real MCP tool calls — your agent does the work, then shows you what it did.

Morning triage

Pull the queue, scored and ready

Show me hot LinkedIn leads from job-change signals this week, grouped by ICP, and surface the top 5 with a one-line reason each.
lead_search lead_get
Agent tuning

Reshape an agent mid-week

Drop SMB roles from my RevOps agent and retarget mid-market North America. Then show me the next run preview.
agent_update agent_runs workspace_icp_options
Outreach drafting

Draft openers grounded in the signal

For my top 3 funding-signal leads, draft 2 connect-note variants each — one pain-led, one congrats-led, both under 280 characters.
lead_enrich lead_draft_outreach
Reporting

Pull weekly numbers without leaving the chat

How many demos came from competitor-engager signals last month vs the month before? Break it out by ICP segment.
lead_search engagement_list workspace_usage
Questions

What people ask before wiring it up

Auth, scopes, supported clients, and how MCP relates to the dashboard. If yours is not here, drop us a line.

Model Context Protocol is an open standard that lets AI agents call typed tools on external services. CatchIntent ships an MCP server so your Claude, Cursor, or Codex agent can operate on your leads, agents, outreach, and workspace — without a custom integration.

Still curious or want a custom scope set?

Email the team
Wire it up

Stop switching tabs. Run pipeline from the AI you already use

Authorize once. Operate leads, agents, outreach, and CRM through the prompt — your dashboard stays open for the visual moments.

28+ tools · OAuth-scoped · Revoke anytime