archestra
Archestra is an all-in-one, open-source enterprise AI platform. You point users, agents and coding tools like Claude Code, Codex or Cursor at one URL, and it provides the pieces an organisation needs to adopt AI safely: an internal chat assistant, a developer portal for LLMs and MCP, gateways for model and tool traffic, and an agent runtime, all built on SSO, role-based access control, sandboxed code execution, guardrails, OpenTelemetry traces and Prometheus metrics.
It suits platform and security teams who need central control over how a company uses models and MCP servers. The LLM gateway supports many providers with cost limits, virtual API keys and dynamic model routing; the MCP gateway uses OAuth with on-behalf-of flows so tools run as the requesting user rather than a shared service account; a private MCP registry and a Kubernetes-based MCP orchestrator let teams publish their own tools; and an A2A gateway, scheduled and webhook-triggered agents, a RAG knowledge base, a mini app builder and deterministic guardrails against risky tool-call patterns round it out. Non-technical users can reach the assistant through Slack, Microsoft Teams or email.
Getting started: pull and run the `archestra/platform` Docker image following the quickstart, then connect model providers, MCP servers and identity settings in the admin interface.
Features
Why choose it
Trade-offs
Maintenance check
- Last commit
- 2026-10-24
- Commits, last 90 days
- 100+
- Releases, last 90 days
- 30
- Latest release
- platform-v1.4.5 (2026-10-24)
- Open issues
- 101
- Recent issues with at least one comment
- 0/2
Oldest week on the left; the latest four weeks are orange. Source: GitHub commit activity.
Status rule: archived on GitHub = Archived; last commit within 90 days = Active; 91 to 365 days = Slowing; older = Stale. All figures come from the GitHub API on the date shown.
Most discussed open issues
- #4145 WhatsApp support for agents similar to MS Teams and Slack25 comments · 2026-04-27
- #3789 KnowledgeBase Document Pruning16 comments · 2026-04-15
- #3784 🚀🚀🚀 WE'RE HIRING9 comments · 2026-04-15
- #3856 Unable to continue A2A session (webhook, slack, msteams) in the chat and vise-versa7 comments · 2026-04-16
- #9427 OpenAPPA: parallel subagent calls leave an orphaned return declaration; next spawn fails with "could not settle a withheld tool call" (503)6 comments · 2026-10-21
Issue titles are quoted from GitHub. Open the link to check whether a problem still applies to the version you use.
Who it fits
- You are a platform or security team that needs central control over how the company uses models and MCP servers.
- You want one platform covering an internal chat, an LLM gateway, an MCP gateway and registry, and an agent runtime.
- You need SSO, RBAC, sandboxed code execution, guardrails, per-user tool authorisation, OpenTelemetry and Prometheus.
- You are a small team that only needs a proxy: Kubernetes-oriented components make this a significant deployment.
- You need to avoid AGPL or commercial terms: the default code is AGPL-3.0 and parts marked Enterprise are under a commercial licence.
- You already run gateways or chat apps whose roles overlap; plan what it would replace.
How it differs
vs toolhive★ 2.3kArchestra is a full AI platform (chat, gateways, runtime) with central policy. ToolHive is narrower: containerised MCP servers, a registry and a gateway, under Apache.
Compatibility
Quick start
Use cases
Alternatives
Featured in these lists
Comments
No comments yet. Be the first!