Archestra

Archestra is open source infrastructure for running enterprise AI in production. It sits between AI clients and agents, such as Claude, Codex, Cursor, n8n and internal agents, and a company's own systems, providing one controlled gateway that combines an LLM proxy, an MCP gateway, identity, policy enforcement, execution, audit and cost control. It is built for platform teams rolling AI out across a company. Identity integrates with Entra ID, Okta, OIDC or SAML, so every tool call is tied to the person who initiated it and user credentials are resolved at runtime. Policies control tool access, approvals, isolation and network egress, and block unauthorized actions and unsafe data flows before execution. Model requests, tool calls, policy decisions and cost are traced and exported through OpenTelemetry and Prometheus. The platform runs locally with Docker, from source, or on Kubernetes with a Helm chart. Archestra is open core. The base platform is licensed under AGPL 3.0, while enterprise features, including role based access control, granular permissions, SSO, a knowledge base with access controlled RAG, data retention windows, content encryption at rest, two factor authentication and idle hibernation for self hosted MCP servers, fall under the Archestra Enterprise License. Companies with fewer than 30 Archestra users in total may use those enterprise features in production at no cost. Above that, an enterprise license is required and pricing is quote based through the sales team, which also offers proof of concept licenses.