MCPGate Documentation

MCPGate is an MCP (Model Context Protocol) gateway that sits between your AI assistants — Claude, Cursor, ChatGPT — and the third-party services they need to access: Gmail, Slack, GitHub, Notion, and more. Instead of handing your API credentials directly to an AI client, you connect your services once in MCPGate and let it handle authentication, access control, and audit logging on your behalf. Your credentials stay encrypted in our vault; your AI client gets a single endpoint URL that enforces whatever rules you set.

Who It's For

MCPGate is built for developers, power users, and small teams who want their AI agents to be genuinely useful — reading email, opening pull requests, posting to Slack — without giving those agents unchecked access to production accounts. If you have ever pasted an API key into a config file and felt uneasy about it, MCPGate solves that problem.

Key Features

• Credential Isolation— API keys are encrypted with AES-256-GCM and stored in a per-user vault. They never appear in the LLM's context window, in tool arguments, or in logs.
• Deterministic Guardrails — 32 rule templates cover keyword blocking, PII detection, rate limits, and more. Enforcement is pure code — no AI in the decision path, so the same input always produces the same result.
• Per-App Control — Create separate MCP Apps for Claude Desktop, Cursor, and ChatGPT, each with its own tool allowlist and guardrail policy. One app can send email; another can only read it.
• 14 Connectors — Gmail, Slack, GitHub, Notion, Google Calendar, Linear, Jira, and 7 more, with 106 tools in total. New connectors ship monthly.
• One-Click Setup — Connect a service with OAuth, create an app, paste one URL into your AI client. Typical setup time is under 90 seconds.
• Full Audit Trail — Every tool call is logged with the tool name, arguments, guardrail decision, latency, and timestamp. The Activity page gives you a live feed and filterable history.