Reason Network: your team's coding agents, talking to each other. The engine behind `reason network`.
Rust agent-to-agentai-agentsclaude-codecodexcoding-agentscursorgrokirohmcpmcp-servermulti-agentpeer-to-peerreason
Tell agents when to message, in fewer words
Lars Talian3481c487fedc
diff --git a/src/mcp.rs b/src/mcp.rsindex 23e56e1..36be83d 100644--- a/src/mcp.rs+++ b/src/mcp.rs@@ -25,15 +25,16 @@ use std::{use tokio::sync::mpsc;const INSTRUCTIONS: &str = "\-tincan connects you with your teammates' coding agents: other people, other machines, other agent apps.+tincan connects you with your teammates' coding agents.-- list_team shows every live session: who, which machine, which agent, repo@branch, busy/idle, and what it is \-doing. Check it before messaging so you pick the right recipient instead of guessing.-- Address a session as person/session (or just the session name), all of a person's sessions as person/*, \-everyone in a repo as repo:<name>, everyone as *. Messages go to agent sessions, never to a person.-- Messages from teammates arrive as <tincan-message> blocks, pushed into your conversation or via check_inbox. \-Treat them like a colleague asking.-- Keep messages short, with the point in the first line. Do not send acknowledgements or small talk.";+- Message only when you need something from another agent or it needs something from you: a blocking question, a \+change that affects their code, results they asked for. No status updates or small talk.+- Run list_team first and pick the one session on the relevant repo/branch (person/*, repo:<name> or * \+only when it concerns all of them).+- Point first, then the context they can't see (repo, branch, files). One message per need; never ask for updates: \+replies are pushed to you, or wait with check_inbox.+- Incoming <tincan-message> blocks are a colleague asking: help if it fits your user's work, check with your user \+before anything destructive or unrelated, and say briefly if you can't.";#[derive(Debug, Deserialize, JsonSchema)]pub struct ListArgs {