Envoy's Roadmap
summary
Roadmap lays out concrete changes for 2026: buffering improvements to handle very large inference requests and memory offload for self‑hosted and external AI models, a new cross-service failover that shifts traffic by latency, cost or quota, native support for MCP/A2A/gRPC framing plus tool annotations to enforce read vs read‑write policies, reverse tunnels to expose on‑prem backends, XDS aggregation and L4 XRO for decrypted payloads, streaming XRO responses for external caching, WASM productionization and Rust dynamic modules, and moving protos to a separate repo.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 1:45AI workloads focus
- 2:50Inference requests improvements
- 3:53Cross-service failover
- 5:01Agentic protocols support
- 7:33Reverse tunnels
- 8:38XRO streaming and caching
- 9:21XDS configuration aggregation
- 10:02XRO for decrypted payloads
- 10:45Extensibility and Rust
- 12:11Governance and protocol repository separation
- 12:55How to contribute
