Orchestrate and manage MCP servers at enterprise scale

Archestra.AI is an open-source, all-in-one enterprise AI platform that puts one URL in front of both human users and agents (including Claude, Codex, and Cursor), then routes behind it to an LLM gateway, an MCP gateway, an agent runtime, and a RAG knowledge base. It ships a chat interface for non-technical staff with Slack, MS Teams, and email front-ends, plus a developer portal that issues a single token for coding assistants like Claude Code, Codex, and Cursor. Enterprise governance is built in rather than bolted on: SSO (OIDC, SAML, Okta, Entra), RBAC with role mapping and team sync, secrets management, and per-environment egress and cost limits.

Reach for Archestra when an organization needs to give many users and agents access to multiple LLM providers and MCP tools under one governed surface, with audit trails, cost limits, and per-user OAuth instead of shared service accounts. It is built for teams already running single-tenant agent tools - the docs specifically call out a migration path from setups like Claude Cowork, OpenClaw, or Hermes - that now need multi-tenant guardrails. It is not a fit for a single developer who just wants a local LLM proxy without the SSO, RBAC, and observability layer, since the platform's value is concentrated in the enterprise controls rather than raw model access.

• LLM gateway for Anthropic, OpenAI, Azure, Bedrock, DeepSeek, and other providers, with cost limits, virtual API keys, and dynamic model routing.
• MCP gateway with OAuth plus On-Behalf-Of authentication, so tools execute as the calling user instead of a shared service identity.
• A2A gateway for agent-to-agent webhook triggers, plus a private MCP registry so teams can publish their own internal tools.
• MCP orchestrator with a Kubernetes operator and self-serve environment promotion.
• Agent runtime supporting scheduled, email, and webhook triggers, sub-agent delegation, reusable skills, sandboxed code execution, and a Kubernetes-native filesystem.
• Deterministic guardrails for tool calls, Dual-LLM verification, and Lethal Trifecta protections against combined data-exfiltration risk.
• Built-in observability: OpenTelemetry traces, Prometheus metrics, logs, and per-team cost tracking.

docker pull archestra/platform:latest

docker run \
  -p 127.0.0.1:9000:9000 -p 127.0.0.1:3000:3000 \
  -e ARCHESTRA_QUICKSTART=true \
  -e ARCHESTRA_BETA=true \
  -v /var/run/docker.sock:/var/run/docker.sock \
  -v archestra-postgres-data:/var/lib/postgresql/data \
  -v archestra-app-data:/app/data \
  archestra/platform

Then open the local dashboard on port 3000. Full Docker, Helm, and Kubernetes deployment instructions - including every environment variable and secrets management setup - are covered in the project's deployment docs. A Terraform provider and a Helm chart are both available for production rollouts.

Platform, security, and infrastructure teams at enterprises that need to open up LLM and MCP tool access to many employees and agents at once, without giving up SSO, RBAC, per-environment cost controls, or an audit trail. Fits organizations already past the single-tenant agent-tool stage and looking for multi-tenant, production-grade guardrails - the docs note three Fortune-50 deployments and $13.5M in total funding as evidence of production readiness. Licensed AGPL 3.0, with a separate Enterprise license tier under an Open Core pricing model.