Learn
What is an agent communication protocol?
An agent communication protocol defines how roles hand off work — message shape, required context, status fields, and escalation paths — so multi-agent systems don't devolve into unstructured chat between bots.
In depth
When three agents collaborate without rules, you get duplicated work, lost context, and circular ping-pong. A communication protocol is the contract between roles: what artifact one role produces, what the next role needs to start, and how failures propagate.
Common patterns on Neuro OS:
- Artifact handoff — research role writes structured brief in git; copy role reads brief id, not raw chat - Status envelope — `queued | in_progress | blocked | done | escalated` on every task - Escalation route — blocked tasks attach reason + suggested human owner - Channel boundaries — support role posts to ticket system; marketing role never writes CRM directly without Ask
Protocols can be encoded in skills and task templates — not improvised each run. This mirrors human teams: Slack messages follow norms; agents need the same discipline.
Examples
- Outbound pipeline — enrich JSON → personalize → approval queue → send
- Incident — triage role tags severity; eng role picks up with linked log excerpt
- Manager role routes intent to sales vs support based on structured intake form
- Blocked handoff — missing CRM field returns explicit error to upstream role
Related terms
FAQ
Is this the same as MCP?
MCP exposes tools. Protocols define how roles coordinate tasks and artifacts above raw tool calls.
Do small teams need protocols?
With one role, minimal. With two+, write the handoff artifact once in a skill.
Run governed agent roles on a company OS — not only definitions in a glossary.