DEV.co
MCP Servers · agentgateway

agentgateway

Agentgateway is an open-source proxy built on MCP and A2A protocols that provides security, observability, and governance for AI agent communication. It routes traffic between agents, LLMs, and tools while enforcing policy, rate limits, and authentication across any framework.

Source: GitHub — github.com/agentgateway/agentgateway
3.7k
GitHub stars
627
Forks
Rust
Primary language
Apache-2.0
License (OSI-approved)

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

FieldValue
Repositoryagentgateway/agentgateway
Owneragentgateway
Primary languageRust
LicenseApache-2.0 — OSI-approved
Stars3.7k
Forks627
Open issues290
Latest releasev1.3.1 (2026-06-22)
Last updated2026-07-08
Sourcehttps://github.com/agentgateway/agentgateway

What agentgateway is

A Rust-based agentic proxy supporting LLM gateway (multi-provider routing with spend controls), MCP gateway (tool federation via stdio/HTTP/SSE), A2A gateway (agent-to-agent), inference routing (K8s integration), guardrails (regex/moderation/webhooks), and observability (OpenTelemetry, JWT/API key/OAuth auth, CEL-based RBAC).

Quickstart

Get the agentgateway source

Clone the repository and explore it locally.

terminalbash
git clone https://github.com/agentgateway/agentgateway.gitcd agentgateway# follow the project's README for install & configuration

Need it deployed, integrated, or customized instead? DEV.co ships production installs.

Best use cases

Multi-LLM Routing & Cost Control

Organizations using multiple LLM providers (OpenAI, Anthropic, Gemini, Bedrock) can unify access through a single gateway, apply budget controls, implement failover, and load-balance across providers without rewriting agent code.

Tool & Data Federation for Agent Workflows

Connect agents to external tools and data via MCP with standardized transports (stdio/HTTP/SSE), OAuth authentication, and OpenAPI integration—enabling complex multi-step workflows without custom integrations per tool.

Agent Governance & Compliance in Enterprise

Deploy agents in regulated environments with fine-grained RBAC (CEL policies), content filtering guardrails (moderation, AWS Bedrock, Google Model Armor), rate limiting, TLS, and full OpenTelemetry observability for audit trails.

Implementation considerations

  • Rust-based binary with no managed SaaS offering; you own deployment, upgrades, and operational responsibility.
  • Requires understanding of MCP protocol and agent communication patterns; not a drop-in proxy for arbitrary protocols.
  • OpenTelemetry integration needed for observability; plan for backend (e.g., Prometheus, Jaeger, Datadog) to collect metrics and traces.
  • RBAC via CEL policy engine requires policy authoring and testing; default-deny posture means explicit policy definition is mandatory.
  • Active development (latest release v1.3.1 on 2026-06-22); expect API surface and behavior changes; review release notes for breaking changes.

When to avoid it — and what to weigh

  • Simple Single-Agent, Single-LLM Setup — If you have a straightforward application with one agent and one LLM provider, the overhead of a dedicated gateway may not justify the complexity. Direct integration may be simpler.
  • Latency-Critical, Sub-Millisecond Inference — Additional proxy hops introduce latency. If sub-millisecond response times are critical and you cannot tolerate an extra network round-trip, direct connections may be required.
  • No Budget for Kubernetes or Production Monitoring — While standalone deployment is documented, the roadmap and community focus appear Kubernetes-centric. Organizations without K8s and observability infrastructure may find setup overhead high.
  • Proprietary Agent Frameworks Without MCP Support — Agentgateway is built on open MCP and A2A protocols. If your agents use proprietary, non-standard communication protocols, integration may require custom work.

License & commercial use

Apache License 2.0 (SPDX: Apache-2.0). Permissive OSI-approved license. Allows use, modification, and distribution for commercial and proprietary purposes under Apache 2.0 terms.

Apache 2.0 explicitly permits commercial use. You may use agentgateway in commercial products, modify it, and redistribute it, provided you comply with Apache 2.0 terms (include license, state changes, provide NOTICE file). No warranty or indemnification; review Apache 2.0 full text for liability disclaimers. No commercial support vendor evident in provided data; requires review of ecosystem for SLAs and support contracts.

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

SignalAssessment
MaintenanceActive
DocumentationStrong
License clarityClear
Deployment complexityModerate
DEV.co fitGood
Assessment confidenceHigh
Security considerations

Agentgateway provides security mechanisms: JWT/API key/OAuth authentication, fine-grained RBAC with CEL policy engine, TLS support, content filtering (regex, moderation APIs, guardrails), and rate limiting. No independent security audit data provided. As a network proxy handling credentials and LLM traffic, treat it as critical infrastructure: isolate in network, audit access logs, rotate credentials regularly, monitor for abuse, and keep dependencies patched. Being Rust-based reduces memory-safety attack surface vs. managed languages, but is not a guarantee of security.

Alternatives to consider

Kong API Gateway + Custom Agent Layer

General-purpose API gateway with plugin ecosystem; requires custom MCP and agent support layers. More mature in operations but not agent-native.

LiteLLM Proxy

Lightweight Python-based LLM routing proxy. Simpler for LLM-only use cases but lacks agent-to-agent, MCP federation, and K8s-native governance features.

Self-Hosted LLM Routing (Custom)

Build your own routing logic in Python/Go/Node. Maximum control but high maintenance burden, no standardized observability or security patterns, and slower time-to-market.

Software development agency

Build on agentgateway with DEV.co software developers

Agentgateway is production-ready for teams building multi-agent systems with policy and cost controls. Review the quickstart, join the Discord community, and evaluate fit for your architecture.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

Related on DEV.co

Explore the category and the services that help you build with it.

agentgateway FAQ

Do I need Kubernetes to use agentgateway?
No. Agentgateway can run as a standalone binary on any OS (Linux, macOS, Windows via WSL). K8s support is optional; docs cover both standalone and K8s deployments.
What LLM providers does agentgateway support?
OpenAI-compatible API is unified endpoint. Docs mention OpenAI, Anthropic, Gemini, and AWS Bedrock. Exact provider list and version compatibility require checking current documentation or source.
Is there commercial support or SLA?
Not stated in provided data. Agentgateway is an open-source project. Community support via Discord and GitHub is available. Commercial support offerings require external review.
Can I use agentgateway with my existing agent framework?
If your framework supports MCP or can be adapted to route via OpenAI-compatible API endpoints, yes. If it uses proprietary non-standard protocols, integration requires custom work or adapters.

Custom software development services

Need help beyond evaluating agentgateway? DEV.co is a software development agency offering software development services and web development for teams of every size. Our software developers and web developers build custom software, web applications, APIs, and mcp servers integrations — and maintain them long-term.

Ready to Build Agent Infrastructure?

Agentgateway is production-ready for teams building multi-agent systems with policy and cost controls. Review the quickstart, join the Discord community, and evaluate fit for your architecture.