Agentgateway is an open-source gateway built for the AI era: one high-performance data plane that routes, secures and observes traditional service traffic, LLM provider traffic, MCP tools and agent-to-agent communication. It is a Linux Foundation project, written in Rust and released under Apache-2.0, with more than 5,400 stars on GitHub and active development through 2026.
As agents multiply, platform teams end up stitching together an API gateway, an LLM proxy and custom MCP plumbing โ each with its own policies and blind spots. Agentgateway unifies them. As an LLM gateway it presents an OpenAI-compatible API in front of OpenAI, Anthropic, Bedrock, Gemini, Vertex and self-hosted models, with credential management, failover, token budgets, semantic caching and prompt redaction. As an MCP gateway it treats tool servers like microservices โ discovering them, signing and scoping every tool call, and keeping an audit trail of what agents invoked. It also routes A2A (agent-to-agent) traffic between frameworks such as LangChain, CrewAI and ADK with identity and tracing on every hop, and does latency- and cost-aware inference routing across vLLM, TGI and Triton replicas. Standard service traffic (HTTP, gRPC, TCP) is covered too, with mTLS and OIDC built in, and an admin UI includes an LLM playground for testing routes.
Agentgateway is completely free and open source under the Apache-2.0 license.
Agentgateway is for platform and infrastructure teams who need one governed, observable front door for all of their agent, MCP and LLM traffic instead of a patchwork of proxies.