MCP1

Product Introduction

1. Definition: MCP1, developed by Appwrite, is a Model Context Protocol (MCP) gateway and enterprise management platform designed to centralize and secure the connection between AI clients (such as Claude, Cursor, ChatGPT, Codex, Gemini, and Windsurf) and backend MCP servers. Operating as a unified HTTP-based proxy endpoint, it handles human authentication, role-based group permissions, per-tool governance, credential sealing, and comprehensive audit logging.
2. Core Value Proposition: MCP1 exists to eliminate the operational friction and security vulnerabilities associated with distributing raw API keys, environment tokens, and server URLs across developer laptops and team infrastructure. By channeling all AI-driven tool calls through a single gateway URL, organizations can enforce strict data governance, monitor real-time AI tool usage metrics, and ensure compliance without sacrificing developer velocity or agent capabilities.

Main Features

1. Unified Gateway Architecture: MCP1 consolidates multiple Model Context Protocol servers behind a single standardized HTTP URL per team or gateway. Clients connect once, handle human authentication via enterprise identity standards, and inherit predefined group scopes without pasting server tokens into local configuration files.
2. Granular Access Control and Policies: Administrators can bundle servers into shared gateways and restrict access using role-based policies. The system supports group-level permissions (full access versus read-only), tool-level filtering, and administrative overrides for destructive operations (such as executing refunds or database drops).
3. Enterprise Observability and Audit Logging: Every tool call, sign-in, and configuration change is captured, measured, and recorded in a centralized audit log. The observability dashboard tracks active usage metrics, success rates, p95 latencies, failure codes, and active human engagement by client and department, with full CSV export capabilities.

Problems Solved

1. Pain Point: Fragmented credential management, unauthorized tool executions, lack of centralized visibility into AI tool consumption, and security risks tied to hardcoding third-party API tokens (GitHub, Stripe, Linear, Notion, Sentry) across local developer environments.
2. Target Audience: Engineering Leads, DevOps Engineers, AI Integration Teams, Chief Information Security Officers (CISOs), and Enterprise Product Organizations seeking secure AI governance.
3. Use Cases: Centralizing multi-departmental AI agent access to internal knowledge bases and production tools; auditing cross-functional LLM queries for compliance; securely granting CI/CD pipelines and automated AI agents scoped access to third-party APIs using expiring, hashed API keys (mcp1_live_*).

Unique Advantages

1. Strengths & Limitations (Pros & Cons):
   • Pros: End-to-end encryption with AES-256-GCM for all server tokens and call arguments; comprehensive compliance frameworks (SOC 2, GDPR, HIPAA, CCPA); one-click catalog integration for over 150 servers; zero-token distribution to end-user machines.
   • Cons: Requires network reliance on the gateway endpoint; introduces an intermediate hop that may add nominal latency (though median latencies hover below 170ms); advanced governance features like IP allowlisting and BAA agreements are locked behind higher-tier enterprise plans.
2. Key Alternatives & Differentiation: Unlike manual, self-hosted open-source MCP proxies that require custom authorization layers and database management, MCP1 provides a managed, production-grade SaaS infrastructure built by Appwrite with native support for enterprise identity, pre-built server catalogs, structured audit exports, and granular rate-limiting out of the box.

Frequently Asked Questions (FAQ)

1. What is an MCP gateway and why do I need one? An MCP gateway is a centralized proxy endpoint positioned between your AI clients (like Claude Code or Cursor) and your Model Context Protocol servers. You need one to secure API credentials, manage team-wide permissions, and maintain an audit log of every tool call made by your AI agents.
2. Where are server tokens stored in MCP1? Server tokens and OAuth grants are securely stored within MCP1, encrypted at rest using AES-256-GCM. Human users never see the workspace's underlying tokens, and credentials are never exposed on local developer laptops.
3. Which AI clients are compatible with MCP1? Any client that supports remote Model Context Protocol over HTTP can connect to MCP1. This includes popular developer tools and AI interfaces such as Claude, Claude Code, Cursor, ChatGPT, Codex, Gemini, VS Code, and Windsurf.
4. How does MCP1 handle security and compliance requirements? MCP1 is fully compliant with SOC 2, GDPR, HIPAA, and CCPA standards. It features encrypted endpoints via TLS, Frankfurt-based data hosting for EU workspaces, and supports Data Processing Agreements (DPAs) along with Business Associate Agreements (BAAs) for Enterprise customers.