Cross-Agent Communication
21 Mar 2026 · Tristan Wright
The Question
Right now, agents working on different users’ projects are isolated. They share hive-scope facts but can’t actively collaborate. Is that enough, or do we need more?
Why This Matters
Imagine Daniel’s agent is building a frontend that calls an API Nathan’s agent is building. Right now, they’d independently discover the API contract through trial and error. Should they be able to share a spec directly?
Constraints
- Cross-agent communication can’t bypass scope boundaries
- Users must be able to see what their agents are sharing
- Communication shouldn’t create dependencies that break when one project changes
Open Questions
- Is hive-scope promotion sufficient for cross-project knowledge?
- Do we need a “shared workspace” concept where two projects can explicitly link?
- Should agents be able to request information from other agents, or is that too risky?
If you have thoughts, raise them in Slack or at the next action day.