# Archestra [Health: Active]

**Category:** 💻 Developer Tools  
**Repository:** https://github.com/archestra-ai/archestra  
**GitHub Stars:** 4391  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/archestra

## Description
Enterprise AI Platform with guardrails, MCP registry, gateway & orchestrator

## Claude Desktop Quick Installation
No confirmed install details yet, so no config is published. Follow the setup steps in the repository above.

## Documentation

## What Archestra MCP server does

Archestra provides a centralized control plane for AI applications, MCP services, LLM providers, and agent workflows. It gives users and developer tools one endpoint for accessing model providers and MCP servers while applying authentication, authorization, policy, and observability controls.

The platform serves both technical and non-technical users. Claude Code, Codex, and Cursor can use its developer portal and proxy, while internal users can interact through chat projects, MCP applications, Slack, Microsoft Teams, or email. Teams can publish internal MCP services through a private registry and promote them across environments.

## How it works

The LLM gateway accepts requests for supported providers, including Anthropic, OpenAI, Azure, Bedrock, and DeepSeek. It can apply virtual API keys, cost limits, and dynamic model routing. The MCP gateway exposes tools with OAuth and on-behalf-of authentication, allowing requests to run under the user's identity instead of a shared service account.

Archestra also includes an MCP orchestrator with a Kubernetes operator and self-service promotion between environments. Its agent runtime supports scheduled, email, and webhook triggers, along with sub-agent delegation, reusable skills, sandboxed code execution, and a Kubernetes-native filesystem. Agent and tool activity can be monitored through OpenTelemetry traces, Prometheus metrics, logs, and per-team cost tracking.

Guardrails can evaluate tool calls deterministically and apply protections for the Lethal Trifecta. The platform adds SSO, RBAC with role mapping and team synchronization, secrets management, environment-specific egress policies, and environment-specific cost limits.

## Setup and configuration

A local quickstart runs the platform as a Docker container:

```bash
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 \
  -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:latest
```

After starting the container, open `http://localhost:3000`. The quickstart mounts Docker access, PostgreSQL data, and application data. Production deployments can use Docker, Helm, or Kubernetes; a Helm chart is recommended for production deployments. Pin an exact image or chart version for production because the `latest` Docker tag generally points to the newest release, which may be a release candidate.

SSO configuration supports OIDC, SAML, Okta, and Entra. Deployment, secrets, identity, access control, and environment behavior are configured through the platform's administration options and deployment settings.

## Tools and capabilities

- MCP gateway with OAuth and on-behalf-of authentication
- LLM gateway for multiple model providers
- Private MCP registry for team-owned services
- MCP orchestration with Kubernetes operator support
- Agent runtime with scheduled, email, and webhook triggers
- Sub-agent delegation and reusable skills
- Sandboxed code execution
- RAG knowledge base connectors
- Tool-call guardrails and Lethal Trifecta protections
- SSO, RBAC, role mapping, team synchronization, and secrets management
- Environment-level egress and cost controls
- OpenTelemetry traces, Prometheus metrics, logs, and team cost tracking
- Mini app builder for chat-based applications

## Limitations and notes

Archestra is distributed under an AGPL 3.0 or Enterprise licensing model. The project maintains stable releases and release candidates; release candidates are built from the main branch and may contain upcoming features. For production use, select a pinned stable image or Helm chart version rather than relying on `latest`.

The platform requires container-based deployment for the documented quickstart and mounts Docker and persistent storage volumes in that setup. Its broad feature set is oriented toward organizations managing shared AI infrastructure, identity, environments, and policy rather than a minimal single-purpose MCP tool server.

_Full upstream README: https://allmcps.com/mcp/archestra/readme_