ArchestackBest of Open-Source AI: List on GitHub ↗ (opens in a new tab)

#4040th of 53 in Gateways and observability

archestra

Self-hosted LLM and MCP gateway with chat, agents, and guardrails

Documentation ↗ (opens in a new tab)Website ↗ (opens in a new tab)Repository on GitHub ↗ (opens in a new tab)
Stars
4.4k
License
custom license
Last commit
Oct 2026
Last release
Oct 2026
Language
TypeScript

Missed the latest GitHub check; these facts may be older.

Overview

Archestra is a self-hosted platform that puts an LLM gateway (Anthropic, OpenAI, Azure, Bedrock, DeepSeek and others) and an MCP gateway with OAuth On-Behalf-Of behind one URL. It adds a private MCP registry, a Kubernetes-based MCP orchestrator, an agent runtime with scheduled, email and webhook triggers, and deterministic tool-call guardrails. Chat front-ends include web, Slack, MS Teams and email, with SSO, RBAC, OpenTelemetry traces and Prometheus metrics.

Who it is for: Platform teams rolling out governed LLM, MCP and agent access across a company

Strengths

  • Covers LLM, MCP and A2A traffic with cost limits and virtual API keys
  • MCP tools run as the calling user via OAuth On-Behalf-Of, not a shared account
  • SSO (OIDC, SAML, Okta, Entra), RBAC with team sync, per-environment egress policies
  • OpenTelemetry and Prometheus built in; Helm chart and Terraform provider available

Weaknesses

  • Open core: AGPL-3.0 plus enterprise licensing, free only for teams under 30 users
  • Quickstart mounts the Docker socket, which grants broad host access
  • The latest tag usually points to a release candidate; production needs pinned versions
  • Broad scope (chat, agents, RAG, app builder) makes it heavier than a plain proxy

What it needs

  • no GPU
  • Needs Docker, PostgreSQL (bundled in quickstart image), Kubernetes (for MCP orchestrator and Helm)
  • Models: Anthropic, OpenAI, Azure, Bedrock, DeepSeek
  • port 3000

Also in Gateways and observability

See all 53
Also in Gateways and observability
RankProjectScore
1OmniRouteLLM gateway routing one OpenAI-compatible endpoint across 372 providers86.8k stars+11,398 (+15%) in 30 daysMIT96 out of 100
2headroomCompresses tool outputs, logs and history before they reach the LLM79.5k stars+4,550 (+6.1%) in 30 daysApache-2.092 out of 100
3new-apiGateway that unifies LLM providers behind OpenAI, Anthropic and Gemini APIs51.8k starsAGPL-3.0known advisory in latest version91 out of 100
#4MLflowTracing, evaluation and prompt registry for LLM apps, plus ML tracking Live demo ↗ (opens in a new tab)28.8k starsApache-2.091 out of 100
#5LiteLLMOpenAI-format gateway and Python SDK for calling 100+ LLM providers63.2k starscustom license88 out of 100