Reason lets Claude Code and Codex assign tasks across computers: they can talk to each other, but still can't automatically prevent code conflicts.

Beating AI News Flash: YC-incubated AI coding company Reason has launched Reason Network, enabling Agents from different vendors such as Claude Code, Codex, Cursor, and OpenCode to communicate directly and assign tasks across computers. For example, you can have Grok Bot assign work to Claude Code on your own computer, and then have Codex on a colleague's computer review the code. Agents send and receive messages directly without humans repeatedly relaying information.

At its core is a real-time updated Agent directory. Developers can see which Agents in the team are online, which repository they are modifying, what branch they are using, and what they are currently busy with. Agents such as Claude Code and Codex can receive messages directly within existing sessions without opening a new chat. Each computer runs a lightweight background program, and devices on the same local area network can communicate directly. Across networks, peer-to-peer connections are prioritized, and when that fails, messages are relayed through encrypted relays.

However, the code conflict avoidance that Reason advertises has not yet been fully implemented. GitHub design documents show that the feature for automatically detecting multiple Agents modifying the same file simultaneously has not yet been developed. At this stage, it mainly relies on sharing work status so that Agents can actively coordinate, and it cannot automatically prevent code from being overwritten.

The underlying communication engine of Reason Network is developed in Rust, derived from the existing open-source project tincan, and uses the MIT or Apache 2.0 license.