Archestra MCP Platform

Content

Archestra is an MCP-native platform for teams that want a private MCP registry, managed MCP gateway endpoints, and runtime governance for shared AI tools. It lets admins approve MCP servers, create installations with personal or team-scoped credentials, assign selected tools to gateway endpoints, and connect clients such as Claude, Cursor, Open WebUI, or custom agents to those curated MCP surfaces.

The project is source-available under AGPL-3.0 and ships a Docker-based quickstart. Its MCP Orchestrator can run self-hosted MCP servers in Kubernetes, while remote MCP servers can be registered and exposed through gateways without Archestra owning their runtime.

Source Review

• https://github.com/archestra-ai/archestra
• https://github.com/archestra-ai/archestra/blob/main/README.md
• https://github.com/archestra-ai/archestra/blob/main/platform/package.json
• https://github.com/archestra-ai/archestra/blob/main/platform/.env.example
• https://archestra.ai/docs/platform-quickstart
• https://archestra.ai/docs/platform-mcp-gateway
• https://archestra.ai/docs/platform-private-registry
• https://archestra.ai/docs/platform-orchestrator
• https://archestra.ai/docs/platform-ai-tool-guardrails
• https://archestra.ai/docs/platform-observability
• https://archestra.ai/docs/platform-deployment
• https://hub.docker.com/r/archestra/platform

These sources were reviewed on 2026-06-06. Prefer the live repository, README, platform package metadata, example environment, quickstart, MCP gateway docs, private registry docs, orchestrator docs, tool guardrail docs, observability docs, deployment docs, and Docker Hub page for current setup, runtime, authentication, and deployment behavior.

Features

• Curate approved MCP servers in a private organization registry.
• Expose selected tools through named MCP gateway endpoints.
• Connect external MCP clients with copied gateway URLs and scoped bearer tokens.
• Install personal or team-scoped MCP connections with static credentials, OAuth, client credentials, enterprise token exchange, or JWKS-based identity.
• Run self-hosted MCP servers in Kubernetes through the MCP Orchestrator.
• Support stdio and streamable-http server transports for self-hosted MCP workloads.
• Assign tools explicitly or resolve credentials at call time based on caller identity.
• Use access control, team visibility, environment restrictions, egress policy, observability, and deterministic tool call/result guardrails.

Installation

Start from the upstream quickstart or deployment docs. For local evaluation, the published image can be pulled with:

docker pull archestra/platform:latest

After Archestra is running, create or install MCP registry entries, assign the approved tools to an MCP gateway, and copy the generated client configuration. A sanitized client configuration looks like:

{
  "mcpServers": {
    "archestra": {
      "url": "LOCAL_ARCHESTRA_MCP_GATEWAY_URL",
      "headers": {
        "Authorization": "Bearer ARCHESTRA_GATEWAY_TOKEN"
      }
    }
  }
}

Review the upstream deployment docs before production use, especially Docker socket access, Kubernetes permissions, secrets storage, identity provider settings, network policy, and gateway token scope.

Use Cases

• Give a team one approved MCP gateway instead of many individual desktop MCP configs.
• Separate approved registry templates from each user's or team's actual credentialed installation.
• Run self-hosted MCP servers in Kubernetes with logs, status, secrets, and restart controls.
• Expose different tool sets for engineering, support, operations, or internal agents.
• Apply deterministic policy to risky tool chains, prompt-injection exposure, and sensitive tool results.

Safety and Privacy

Archestra centralizes MCP access, which makes its admin and runtime boundaries important. The quickstart's Docker socket mount gives the platform privileged control over the host Docker daemon. Use an isolated evaluation host, review container permissions, and prefer hardened Kubernetes deployment patterns for shared environments.

Credential resolution is a major part of the platform. Be explicit about personal versus team-scoped installs, gateway visibility, bearer token scope, header passthrough, OAuth refresh, external identity exchange, and network egress. Observability is useful, but logs and traces can include sensitive MCP tool metadata, prompts, arguments, results, user IDs, and team context.

Duplicate Check

No archestra-ai/archestra entry, Archestra MCP Platform entry, Archestra MCP gateway entry, or matching source URL was found in content/mcp.

• Source repositorygithub.com 2026-07-21T07:17:01+00:00
• Documentationarchestra.ai
• Submitted by oktofeesh12026-06-06