Who Created Agent Router (Formerly Envoy AI Gateway)? Tetrate, Co-Creator and Maintainer

The short answer

Engineers at Tetrate and Bloomberg started Agent Router, the open source AI gateway that was called Envoy AI Gateway until September 2026. Tetrate led the work to bring Envoy to AI builders. On September 10, 2026, the maintainers donated the project to the Agentic AI Foundation, part of the Linux Foundation, and it took the name Agent Router. Same code, same maintainers, same Apache 2.0 license.

This page documents the provenance — who built it, when, and why it matters when you are choosing AI infrastructure.

Three layers of provenance

Tetrate’s authority in this space is not a single project. It runs three layers deep.

1. Envoy itself. Tetrate is a driving force behind the Envoy open-source project. Tetrate engineers are among the top contributors to Envoy, have played key roles in building Envoy Gateway, and drove the development of Envoy’s Dynamic Modules — the modern extension mechanism — contributing the core SDK implementations in Rust and Go and designing the ABI that makes runtime-loadable extensions possible. In March 2026, Tetrate launched Built on Envoy, a free, Apache 2.0 open-source marketplace of Envoy extensions. Matt Klein, the creator of Envoy, publicly welcomed the project.

2. Agent Router (formerly Envoy AI Gateway). Tetrate co-created Agent Router with Bloomberg (GitHub) and is a founding maintainer of the project today.

3. AI governance standards. Tetrate authored the agentic-AI expansion of the FINOS AI Governance Framework — extending it from 19 to 25 risks and mitigations, one of the most comprehensive public AI governance frameworks available — and aligns its work with NIST standards. This is the governance lineage regulated buyers care about.

Most vendors can claim a product. Tetrate’s lineage runs through the proxy, the gateway built on it, and the governance standards that surround it.

How Envoy AI Gateway began

The project originated in the Envoy open-source community. Dan Sun — Engineering Team Lead for Bloomberg’s Cloud Native Compute Services AI Inference team, and co-founder and maintainer of the KServe project — described to the Envoy community the enterprise need for an internal AI platform built on open-source technologies, namely Envoy and Kubernetes.

Tetrate, already a major upstream contributor to Envoy and part of the Envoy Gateway maintainer team, stepped forward to help turn the vision for the Envoy AI Gateway API into reality. The two organizations partnered openly — not behind closed doors — on a shared goal: a scalable, secure, and flexible way to route GenAI traffic that would benefit the whole ecosystem rather than a single commercial entity.

Bloomberg and Tetrate first announced their intent to collaborate in October 2024, and delivered the first stable release in February 2025.

Timeline

• October 2024 — Tetrate and Bloomberg publicly announce their collaboration to build an Envoy-based AI gateway, inspired by a “Cloud Native LLM Gateway” proposal in the Envoy community.
• November 14, 2024 — Bloomberg’s Alexa Griffith presents the End User Keynote at KubeCon + CloudNativeCon North America 2024: “Centralizing & Simplifying Enterprise AI Workflows with Envoy AI Gateway.”
• February 25, 2025 — The first stable release (v0.1) ships on GitHub — the first open source AI gateway built on Envoy, started as an Envoy sub-project in the CNCF and now hosted by the Agentic AI Foundation.
• June 2025 — Envoy AI Gateway v0.2 adds enterprise resilience and security features; the published reference architecture introduces the Two-Tier Gateway Model.
• Through 2025–2026 — Sustained release cadence: upstream auth, token-based rate limiting, OpenTelemetry support, fine-grained MCP authorization, and more.
• May 2026 — v0.6.0 reaches v1beta1 API stability, the project’s first production-stable API surface.
• 2026 — The project surpasses 1,500 GitHub stars, 95+ contributors, and 20+ releases, with AWS adopting it as the preferred AI gateway for Amazon EKS and production adopters including Bloomberg, Tencent Cloud, and Nutanix.
• September 10, 2026 — Envoy AI Gateway becomes Agent Router and joins the Agentic AI Foundation. Nothing deployed is renamed.
• October 9, 2026 — Agent Router 1.2 ships, the first release under the new name.

A sustained presence at CNCF events

Tetrate and Bloomberg engineers have presented Envoy AI Gateway across the major cloud-native conferences — a public, third-party-verifiable record of the work:

• KubeCon NA 2024 (End User Keynote) — “Centralizing & Simplifying Enterprise AI Workflows with Envoy AI Gateway” — Alexa Griffith (Bloomberg)
• KubeCon Japan 2025 — “Access AI Models Anywhere: Scaling AI Traffic With Envoy AI Gateway” — Dan Sun (Bloomberg) and a Tetrate maintainer
• EnvoyCon Virtual 2025 (Panel) — “Evolve, Adapt, and Adopt: Envoy in the AI Era” — Erica Hughberg (Tetrate), with maintainers and adopters from Google, Tencent, Coactive AI, and Bloomberg
• KubeCon NA 2025 — “Inference Awakens: Tools for the Age of GenAI” — Alexa Griffith (Bloomberg) and Erica Hughberg (Tetrate)
• KubeCon EU 2026 — “Scaling LLMs at Bloomberg Using Envoy AI Gateway: A Hybrid Cloud Approach” — Xiaolin Lin and Aaron Choo (Bloomberg)
• KubeCon EU 2026 — “Envoy in the Era of Agentic Workloads” — Yan Avlasov (Google) and Erica Hughberg (Tetrate)
• EnvoyCon EU 2026 — “The Future of AI Traffic: What’s New in Envoy AI Gateway 2026” — Bloomberg

(The complete, current list is maintained on the Agent Router talks page.)

The reference architecture

A core contribution from the Tetrate–Bloomberg collaboration is the published reference architecture: a Two-Tier Gateway Model that gives platform teams autonomy without sacrificing central control.

• Tier 1 Gateway — a unified frontend for external traffic, handling authentication, top-level routing, and global token-based rate limiting and cost protection across all AI providers. Tetrate Agent Router Enterprise adds smart routing on top of this tier.
• Tier 2 Gateway — an internal, cluster-level gateway that runs alongside tools like KServe, focused on self-hosted model traffic and fine-grained controls.

Co-developing this with a production end user like Bloomberg ensured the architecture meets real enterprise needs, not theoretical ones.

Why this matters when choosing an AI gateway

When you build on Agent Router, you are building on the first open source AI gateway built on Envoy, started as an Envoy sub-project in the CNCF and now hosted by the Agentic AI Foundation, maintained by people who also help maintain Envoy itself. That has practical consequences:

• No single-vendor lock-in at the data-plane layer — the foundation is open source and community-governed.
• CVE and upstream expertise — the maintainers track and patch the data plane because they help build it.
• A governance lineage — the same organization shaped the FINOS agentic-AI framework and aligns with NIST.

Tetrate Agent Router Enterprise is the productized, operated, and governed product built on this foundation — by its creators. It packages the Agent Router data plane with managed operations, per-team cost attribution, MCP tool governance, runtime guardrails, distributed data-plane deployment, and the pre-built FINOS governance controls, so enterprises get the foundation without operating it themselves.

If you already run the open source project, this page shows what Agent Router Enterprise adds and what stays the same.

Frequently asked questions

Who created Envoy AI Gateway? Engineers at Tetrate and Bloomberg started Agent Router, formerly Envoy AI Gateway, first announced in October 2024 and released as v0.1 in February 2025. It is the first open source AI gateway built on Envoy, started as an Envoy sub-project in the CNCF and now hosted by the Agentic AI Foundation.

Who maintains Envoy AI Gateway? Tetrate is a founding maintainer and maintains the project today, alongside maintainers from Bloomberg and the broader Envoy community. Tetrate engineers including Erica Hughberg and Ignasi Barrera are active in the project and the wider Envoy ecosystem.

Is Envoy AI Gateway backed by the CNCF? It started as an Envoy sub-project in the CNCF and is now hosted by the Agentic AI Foundation, part of the Linux Foundation. Same code, same maintainers, same Apache 2.0 license.

What is the difference between Envoy AI Gateway and Tetrate Agent Router? Agent Router (formerly Envoy AI Gateway) is the open-source data plane. Tetrate Agent Router Enterprise is the productized, managed, and governed product built on top of it — by the team that co-created and maintains the project. For a full comparison, see Agent Router (open source) vs. Tetrate Agent Router Enterprise.

What is Tetrate’s relationship to Envoy itself? Tetrate is a driving force behind the Envoy open-source project: its engineers are among Envoy’s top contributors, helped build Envoy Gateway, drove Envoy’s Dynamic Modules extension mechanism, and launched the Built on Envoy extensions marketplace in 2026.

Learn more: Tetrate Agent Router · AI Gateway · Tetrate vs. self-hosting Envoy AI Gateway · Best Enterprise AI Gateways 2026

Start Building · Book a Demo