Connect your AI agent over MCP
Chuhching is agent-native: your AI agent can drive the whole platform through MCP (the open standard for giving agents real tools). But raw tools only give an agent capability — which tool to use, in what order, with what guardrails is judgment. The Chuhching skills pack adds that judgment, so your agent uses Chuhching like a trained operator instead of guessing. This guide gets both installed in a couple of minutes.
The two pieces
- The MCP connection — the tools. Authenticated with a scoped access token you generate in Chuhching.
- The skills pack — twelve skills that encode the right workflows and guardrails: an orchestrator plus getting-started, CRM & pipeline, inbox & outreach, automations, wiki/second-brain, and opt-in modules for podcast guesting, PR, scheduling, community & affiliates, social, and game launches.
Why the skills, not just the MCP
The skills carry the things a raw tool list can't: that deal values are in cents, that you list a pipeline's stages before moving a deal, that outreach is checked for CAN-SPAM and SMS compliance before anything sends, that social posts are drafted (never auto-published), and that paid discovery is quoted before it spends. That is why the same task comes out right every time instead of occasionally.
Install in Claude Code
- Generate a token. Open Settings → Integrations → MCP and create a scoped access token (starts with
piq_pat_). Read is always on; add write scopes (outreach, inbox, pipeline, scheduling) as needed. - Download the skills. Download the skills pack (.zip) and extract the
skills/folders into~/.claude/skills/(available in every project) or a project's.claude/skills/. Skills are picked up automatically — no registration step. - Set your token as the
CHUHCHING_PATenvironment variable — in your shell before launching, or in Claude Code's.claude/settings.local.jsonunderenv. It is read at runtime and never stored in any file you share. - Add the MCP server. Put this in
~/.claude/.mcp.json(or a project.mcp.json):{ "mcpServers": { "chuhching": { "type": "http", "url": "https://mcp.chuhching.com/mcp", "headers": { "Authorization": "Bearer ${CHUHCHING_PAT}" } } } } - Restart and go. Just tell your agent: “Onboard me to Chuhching.” It will set up your profile, wiki, pipeline, and first targets.
Other agents (Cursor, Windsurf, Cline)
Use the same skills pack — extract it into that agent's skills directory, then add the Chuhching MCP server in the agent's own MCP settings using the config above. Off Claude Code the skills and MCP are added separately (once each).
What your agent can now do
Ask in plain language: find shows to pitch, draft and send an on-brand pitch, reply to your inbox, create and move deals, score a lead, build an automation (“when a deal is won, tag it and send the welcome email”), search your second-brain, draft social posts, or plan a game launch. The right skill fires on intent.
Safety
- Tokens are scoped and stored only as a hash — revoke and regenerate anytime.
- The skills confirm outward actions (sending, publishing, booking, spending) before acting.
- Compliance is enforced before any email/SMS goes out; social is draft-only.
- Paid discovery (podcast/curator/creator search) is quoted before it spends.
Next
Prefer marketing context first? See the Connect your AI agent overview. To manage or revoke tokens, go to Settings → Integrations → MCP.