agentgateway
Agentgateway is an open-source proxy built on AI-native protocols, MCP and A2A, which the README calls the first complete connectivity solution for agentic AI. It combines several gateways in one: an LLM gateway that routes traffic to major providers (OpenAI, Anthropic, Gemini, Bedrock and more) through a unified OpenAI-compatible API with budget and spend controls, prompt enrichment, load balancing and failover; an MCP gateway that connects LLMs to tools and data with tool federation, stdio, HTTP, SSE and Streamable HTTP transports, OpenAPI integration and OAuth; and an A2A gateway for secure agent-to-agent communication with capability discovery and task collaboration.
It also offers inference routing to self-hosted models through Kubernetes Inference Gateway extensions, with decisions based on GPU utilisation, KV cache, LoRA adapters and queue depth, multi-layer guardrails (regex, OpenAI moderation, AWS Bedrock Guardrails, Google Model Armor and custom webhooks), and security and observability features: JWT, API keys and OAuth authentication, fine-grained RBAC with a CEL policy engine, rate limiting, TLS and OpenTelemetry metrics, logs and tracing. A standalone quickstart is provided, and the code is Apache-2.0 licensed.
Features
Why choose it
Trade-offs
Maintenance check
- Last commit
- 2026-10-23
- Commits, last 90 days
- 100+
- Releases, last 90 days
- 9
- Latest release
- v1.7.0-alpha.1 (2026-10-21)
- Open issues
- 290
- Recent issues with at least one comment
- 2/14
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
- #2607 Follow-on standalone configuration importer features20 comments · 2026-07-20
- #3230 docs: add observability platform integration guides14 comments · 2026-08-28
- #2696 Configuration validation failed14 comments · 2026-07-27
- #239 Feature Request: Enterprise SSO Integration with Per-User Authentication to External MCP Services15 comments · 2025-07-30
- #3505 [Discussion]: Add AGENTS.md and skills for contributor workflows8 comments · 2026-09-16
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 want one proxy for LLM calls (OpenAI-compatible API with spend controls, failover and load balancing), MCP tools and A2A traffic.
- You need guardrails from several sources: regex, OpenAI moderation, AWS Bedrock Guardrails, Google Model Armor or webhooks.
- You need security and observability built in: JWT, API keys, OAuth, RBAC with a CEL policy engine, rate limiting, TLS and OpenTelemetry.
- You only need to bridge one MCP transport: a small proxy is simpler.
- You want the inference-routing features without Kubernetes: the README ties them to Kubernetes Inference Gateway extensions.
- You want detailed setup instructions in the README: it is short and points to the documentation site.
How it differs
vs mcp-context-forge★ 4.7kAgentgateway is a proxy that also covers LLM routing, guardrails and spend limits. ContextForge is a gateway and registry that federates and wraps services, with an admin UI and multi-tenancy.
vs toolhive★ 2.3kToolHive focuses on running MCP servers in locked-down containers with a registry and audit logs. Agentgateway focuses on traffic: LLM, MCP and A2A routing, policy and guardrails.
Compatibility
Quick start
Use cases
Alternatives
Featured in these lists
Comments
No comments yet. Be the first!