Cloudflare Gives Every AI Agent Its Own Computer: @cloudflare/computer, Cross-Language RPC, and Inbound TCP Land on Day One of Agents Week

Cross-Language RPC: Python and JavaScript Workers Can Now Talk Directly

Also on Monday, Cloudflare announced that Python Workers and JavaScript/TypeScript Workers can now call each other's methods directly via Workers RPC using Service bindings . The cross-language calls work without extra dependencies, schema definitions, protobuf, or serialization code — Pyodide's Foreign Function Interface (FFI) handles type conversion automatically, and live objects, functions, and streams can cross the language boundary . Exceptions propagate to the call site, and structured cloneable types are supported as parameters or return values .

Inbound TCP and gRPC Support via Spectrum

Cloudflare Workers and Containers now support inbound TCP connections through Spectrum, allowing direct socket forwarding to Durable Objects and Containers . A new connect(socket) handler in the Workers runtime lets a Worker directly accept an inbound TCP socket . Developers can run full-duplex gRPC applications or use automatic gRPC-to-gRPC-web translation directly within Workers — opening the Workers platform to protocols beyond HTTP for the first time .

What These Announcements Mean for the Agent Economy

Taken together, these three announcements form a coherent platform strategy:

• @cloudflare/computer solves the fundamental "where does my agent live and work?" problem — giving agents a persistent, stateful computing environment with the right execution model per task, rather than forcing everything into a container or a serverless function .
• Cross-language RPC lets agent developers compose multi-language systems without friction: a Python agent can call a JavaScript data-processing Worker or vice versa, treating the whole Workers edge network as one seamless runtime .
• Inbound TCP and gRPC means agents and the services they talk to are no longer limited to HTTP — they can use streaming, bidirectional, and low-latency protocols that real-time agentic workloads often require .

The collective message is that Cloudflare is building the infrastructure layer for the agent era: agents get their own computers (not just containers), languages become interchangeable, and network protocols stop being a constraint. This positions the platform as an end-to-end environment where developers can build, deploy, and interconnect autonomous agents at the edge without managing servers, containers, or protocol translation.