AI systems are starting to look a lot more like distributed systems.
Agents are calling tools. Models are routing across providers. Workflows are spanning APIs, MCP servers, databases, and other agents. As these systems move from experiments into real operational environments, organizations are quickly realizing that the infrastructure patterns built for traditional web traffic do not fully address the governance, observability, routing, and security challenges agentic systems introduce.
That is the problem the open source project agentgateway was built to solve.
To help the community address these emerging infrastructure challenges, agentgateway is joining the Agentic AI Foundation (AAIF) as the fourth hosted project under the Linux Foundation.
What is agentgateway
Agentgateway is an open source gateway built for the way modern AI systems actually work. It provides a unified layer for managing MCP traffic, Agent-to-Agent communication, LLM inference, REST APIs, and gRPC services through a single operational surface.
Instead of standing up separate infrastructure for AI systems, organizations can use agentgateway to manage AI and traditional application traffic together using the same security controls, observability pipelines, routing policies, and governance models.
For platform teams, this means agent traffic can be managed with familiar operational primitives instead of requiring a separate, AI-only infrastructure stack. The same kinds of controls teams already care about - security, observability, routing, governance, and reliability - can apply to agent workflows too.
Key capabilities include:
- MCP and A2A support: routing, federation, and traffic management for modern agent interoperability protocols
- Model independence: seamless switching between LLM providers, with native flexibility for open-weights models
- Unified data plane: one gateway for HTTP, gRPC, LLM inference, and agent traffic
- Security controls: JWT authentication, API key auth, RBAC, external authorization, mTLS, CORS, and protections against malicious tool behavior
- Built-in observability: metrics, tracing, and access logging designed for AI and agent workflows
- MCP virtualization: federate multiple MCP tool servers into a single access point with the ability to configure tool-level access policy for MCP clients.
- Declarative policy with CEL: dynamic configuration and extensibility using Common Expression Language (CEL)
- Model and tool governance: rate limiting, content-based routing, prompt guards, budget controls, and model aliasing
- Platform-agnostic deployment: runs on bare metal, VMs, containers, and Kubernetes, with full Gateway API conformance
- Dynamic configuration: updates through xDS without downtime
- High-performance architecture: built in Rust for low-latency, high-throughput workloads
Why AAIF
AAIF provides an open and neutral home for critical agentic AI standards, protocols, and open source projects, with governance hosted by the Linux Foundation.
That includes standards like MCP, open source tooling like goose, and infrastructure projects like agentgateway that help organizations operate AI systems safely and reliably at scale.
Agentgateway strengthens the ecosystem by providing a shared infrastructure layer for managing AI traffic, agent workflows, interoperability, security, and governance across tools and services. It gives the community a place to collaborate on the operational foundations agentic systems will need as they become more widely adopted.
"The addition of agentgateway marks a major milestone for the AAIF ecosystem. As organizations deploy increasingly sophisticated AI systems, they need infrastructure that can provide visibility, governance, and operational control across agent workflows and tool interactions. agentgateway helps fill that gap with an open, high-performance platform designed specifically for these emerging workloads." — David Soria Parra, Chair, AAIF Technical Committee
"Donating agentgateway to AAIF is the natural next step for a project built around open connectivity and interoperability. Our goal from day one has been to help organizations manage the operational realities of AI systems — security, governance, observability, and reliability — through open infrastructure. Bringing the project into AAIF ensures it can evolve through open collaboration and neutral governance alongside the broader ecosystem." — Lin Sun, Head of Open Source, Solo.io & agentgateway Contributor
Agentgateway joins a growing AAIF ecosystem focused on making agentic systems more interoperable, manageable, and trustworthy.
The project is licensed under Apache 2.0, and the community is already active with 300+ active contributors across over 60 organizations, including CoreWeave, Red Hat, Solo.io, Adobe, Salesforce, Amdocs, and Microsoft.
Get involved
- Explore the docs
- Star and contribute on GitHub
- Follow the get started guide
- Join the conversation on Discord
- Attend our weekly community meetings
* Thanks to the AAIF, where this blog was originally posted: https://aaif.io/blog/agentgateway-joins-aaif-as-an-open-gateway-for-agentic-ai-infrastructure/
%20a%20Bad%20Idea.png)
%20For%20More%20Dependable%20Humans.png)
