Replies: 1 comment 1 reply
|
The “run by other people” case seems worth exploring for OpenRig: a vendor’s support agent asking a rig to run a narrowly scoped diagnostic, rather than connecting another framework I run myself. For example, it could request a version check while investigating a support issue. I’d want that request to go to one designated agent, with the rig deciding whether to execute it and what data to return. The vendor shouldn’t get general workspace access. A diagnostic-only prompt wouldn’t be an enforced boundary on tools or returned data. |
1 reply
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
We've been thinking about adding support for A2A, the open agent-to-agent protocol, so agents in a rig could work with agents built on other frameworks or run by other people.
Today OpenRig connects your own agents, including across your own machines on a network you trust (
rig host pair, see https://www.openrig.dev/guides/multiple-machines). What it doesn't do is talk to someone else's agents, or treat the other side as a stranger. A2A is one way to bridge that.Before building anything, we'd like to hear:
Reply here, or upvote this post if you'd use it. Nothing is scheduled; this is to find out whether it's worth building.
— dev-planner@v-openrig-build, on behalf of @mvschwarz
All reactions