
One URL for every MCP server
MCP1 by Appwrite is a hosted Model Context Protocol (MCP) gateway for teams. It puts every MCP server a team uses behind a single URL (e.g. mcp1.com/mcp/moonbase/engineering), so humans authenticate once with their own account while admin-defined server tokens stay sealed in the gateway instead of being pasted onto individual laptops. Admins install servers from a one-click catalog or by URL, bundle them into shared gateways assigned to groups (full or read-only access, plus per-tool rules), and every tool call is checked against policy and recorded in an audit log. It also manages agent context: instead of front-loading all tools, the gateway exposes a tool search and delivers server instructions lazily, keeps unreachable tools out of context entirely, and trims oversized tool responses. It ships with a dashboard of usage analytics (tool calls, success rate, p95 latency per server, active humans by group and client), CI keys scoped to a single gateway with expiry, and compliance posture including SOC 2, GDPR, HIPAA (BAA on Enterprise), CCPA, AES-256-GCM encryption and EU hosting in Frankfurt.
Admins install catalog servers in one click (or add any remote MCP server by URL) and bundle them behind a single gateway URL per team; clients only ever see one URL and MCP1 handles tokens, health checks and tool syncing.
Grant shared gateways to groups fully or read-only, restrict specific servers or individual tools to certain groups, hold members to read-only, and require admin approval for destructive tools.
Every tool call, sign-in and workspace change is recorded with actor, client, gateway, latency and result (allowed/blocked/failed), with filtering, live view and CSV export.
The gateway starts agents with a few lines, then delivers tools via tool search (mcp1__search_tools), server instructions on first use, only tools the caller can actually use, and truncated answers with a note to narrow the query.
Admins see workspace-wide usage across 7/30/90-day windows (tool calls, success rate, p95 latency, active humans by group and client) to spot failing or slow servers, measure AI adoption and decide which servers to keep; members see their own usage.
Every member can build a personal gateway combining their own tokens with workspace servers open to their groups; scoped, expiring (e.g. 90-day) API keys let pipelines and headless agents call a single gateway without a browser.
273 servers across categories like Engineering (107), Marketing (63), Data & analytics (54), Operations & productivity (49), Product & design (35), Finance & payments (35), Sales & CRM (31), Knowledge & docs (32), Customer support (12) and Security & IT (16), plus anything else addable by URL.
SOC 2, GDPR (with a DPA), HIPAA (BAA on Enterprise) and CCPA; server tokens, OAuth grants and call arguments sealed with AES-256-GCM, TLS on every endpoint, EU hosting in Frankfurt, encrypted backups, and instant access revocation when someone is removed from the workspace.
Admins install MCP servers from the catalog or by URL (tokens stay sealed in MCP1), bundle them into one gateway URL per team, and grant that gateway to groups with full or read-only access, down to per-tool rules.
Users add the gateway URL to their MCP client (Claude, Claude Code, Cursor, ChatGPT, Codex, Gemini, etc.) and sign in with their own account; their group membership determines which servers and tools they can call.
Each tool call is checked against the caller's group policies, then logged with actor, client, gateway, latency and result; admins get full observability and can export the audit log as CSV.
Free for individuals, per-member pricing for teams. Individual is free (1 member, 3 gateways, 15 servers, all catalog servers with one-click auth, policies and audit log). Team is $12 per member/month (up to 25 members, 25 gateways, 100 servers, groups, shared gateways, per-tool rules, usage analytics and audit export). Enterprise is $29 per member/month (up to 1,000 members, 500 gateways, 1,000 s
Free
For one human wiring up their own agents: 1 member, 3 gateways, 15 servers, every catalog server with one-click auth, policies and audit log.
$12 per member / month
For teams sharing gateways with groups: up to 25 members, 25 gateways, 100 servers, groups, shared gateways and per-tool rules, usage analytics and audit export.
$29 per member / month
For organizations running MCP at scale: up to 1,000 members, 500 gateways, 1,000 servers, everything in Team plus IP allowlists for gateways and invoiced billing on request.
MCP1 is an MCP gateway for teams that puts every Model Context Protocol server your team uses behind one URL, with authentication, group-based access, per-tool rules and an audit log. It works with Claude, Cursor, ChatGPT, Codex, Gemini and any MCP client.
Humans authenticate once with their own account when adding the gateway URL to their client. Server tokens are sealed inside MCP1 and never sit on individual laptops. For pipelines and agents without a browser, MCP1 issues API keys made for one gateway, scoped to specific servers and set to expire, stored only as a hash.
The site lists Claude, Claude Code, Cursor, ChatGPT, Codex and Gemini, and states it works with any MCP client — clients just add the gateway URL and sign in.
Yes. The catalog has 273 servers installable in one click, and you can add any remote MCP server by URL. Personal gateways can also include servers with your own tokens alongside workspace servers open to your groups.
Instead of front-loading all tools, the gateway starts agents with a few lines about itself, exposes tool search (mcp1__search_tools) to find tools from a few words, delivers each server's instructions on first use, hides tools and servers the caller can't use, and trims oversized tool answers with a note to narrow the question.
Individual is free for one person (3 gateways, 15 servers). Team is $12 per member/month (up to 25 members, 25 gateways, 100 servers). Enterprise is $29 per member/month (up to 1,000 members, 500 gateways, 1,000 servers, with IP allowlists and invoiced billing on request).
MCP1 advertises SOC 2 (audited controls for security, availability and confidentiality), GDPR with a data processing agreement, HIPAA safeguards with a BAA available on Enterprise, and CCPA. Data is encrypted with AES-256-GCM (tokens, OAuth grants and call arguments), TLS on every endpoint, hosted in Frankfurt, with encrypted backups and a tested business continuity plan.
Yes. Every call, sign-in and change is recorded in an audit log with actor, client, gateway, latency and result. Admins see workspace-wide analytics (tool calls, success rate, p95 latency per server, active humans by group and client) and can export the audit log as CSV; members see their own usage.