MCP

Model Context Protocol (MCP) is an open standard for connecting AI assistants to tools and data. Onboard exposes MCP so your team can use operational AI for onboarding copilots, launch coordination, and support workflows—not only generic chat.

What you get

• Intent-based workflow tools — Search customers, list active projects, summarize onboarding status, load KPI dashboards, find blockers, draft updates, and preview writes—without hand-authoring REST calls.
• Curated [API] tools — Full OpenAPI browse, explain, and run_api_request when you need raw routes.
• Hosted MCP (HTTP) for clients that support remote MCP over HTTP, using long-lived sessions (for example Server-Sent Events) and JSON-RPC as documented on MCP setup.
• OAuth or API key — Claude custom connectors can sign in per user; most desktop clients use an API key from Integrations → API.
• Optional local stdio server for engineering workflows (for example running python -m onboard_mcp.server in a secure environment). Email [email protected] to confirm whether this is enabled for your tenant.

Security and access

MCP calls are only as safe as the credentials and network you use. Treat MCP like any privileged integration surface:

• Use API keys and secrets only where your security policy allows, and rotate them on the same schedule as other production keys.
• Prefer server-side or controlled desktop setups over pasting keys into untrusted tools.

For broader trust and compliance context, see Compliance & trust. For operational checks before go-live, see the Security checklist. For how API keys work with Onboard, see Authentication.

Learning path (curriculum)

0. MCP for customer teams — features, what teams do day to day, rollout (start here for business stakeholders)
1. Introduction to MCP — what MCP is and how Onboard uses it
2. MCP setup — production URLs, API key, authentication
3. Connecting to Onboard MCP — how to connect clients, OAuth vs API key, MCP Inspector, Client ID from Integrations
4. Security & permissions (MCP) — keys, tenancy, reviews
5. AI workflow examples — daily brief, QBR prep, follow-up, launch week
6. MCP use cases — AI agents — Chief of Staff, set-up, onboarding, review, reporting, opportunity, and risk agents
7. Agent MCP playbook (engineers): in the API repo, docs/api/mcp-agent-tool-playbook.md and .cursor/skills/onboard-agent-* skills map each persona to workflow and [API] tools (including browse_inbox, list_projects, get_kpi_dashboard, and post_comment).
8. Example workflows — leadership review, launch coordination, meeting follow-up
9. Automation platforms & MCP — Zapier, Make, n8n, Workato vs MCP
10. MCP desktop clients — per-client setup guides (Claude, Quick, Cursor, …)

Setup highlights

• MCP on Amazon Quick — field-by-field Quick configuration with a screenshot.
• MCP on OpenAI Agent Builder — MCP tool node, + Server, and API key auth for workflows.

Related docs

• Security checklist — keys, webhooks, MCP hygiene
• AI assistant rules — copyable rules bundles for Cursor, Claude, Windsurf, and Paperclip when building against the REST API.
• Onboard REST API — base URLs, auth, and API behavior that underpin MCP-backed tools.