Pre-checks
What problem does this solve?
Agents need a persistent way to understand their relationships with other agents and humans.
Today, an agent may participate in workflows involving managers, collaborators, reviewers, owners, or peer agents, but the relationship context is not treated as durable product state that the agent can reliably recognize later. If relationships are only implied by a conversation, an assignment, or a temporary runtime context, the agent may fail to understand who it is talking to, who it should coordinate with, or what role another participant plays.
This is important for multi-agent collaboration, approval flows, handoffs, and human-agent teamwork.
Proposed solution
Add persistent relationship modeling between agents and between agents and humans, and expose that relationship context to agents when they operate.
Expected behavior:
- Persist relationships such as owner, manager, collaborator, reviewer, delegate, peer, customer, or custom relationship types.
- Support relationships between agent-to-agent, human-to-agent, and potentially human-to-human if needed for agent context.
- Make relationships visible and editable in an appropriate admin/settings surface.
- Include the relevant relationship context in the agent's runtime prompt/context so the agent can autonomously recognize the participant's role.
- Allow agents to use this relationship context when deciding how to route, escalate, delegate, or respond.
- Keep relationship state auditable so users can understand why an agent treated another participant in a certain way.
Suggested acceptance criteria:
- A relationship configured once remains available across sessions and restarts.
- An agent can identify the role/relationship of another agent or human during later interactions.
- Users can view and update the relationship map from the UI.
- Relationship context is scoped by tenant/company and does not leak across organizations.
- Existing agent relationship features, if any, are reconciled with this persistent model instead of creating a parallel temporary mechanism.
Willing to contribute?
Pre-checks
What problem does this solve?
Agents need a persistent way to understand their relationships with other agents and humans.
Today, an agent may participate in workflows involving managers, collaborators, reviewers, owners, or peer agents, but the relationship context is not treated as durable product state that the agent can reliably recognize later. If relationships are only implied by a conversation, an assignment, or a temporary runtime context, the agent may fail to understand who it is talking to, who it should coordinate with, or what role another participant plays.
This is important for multi-agent collaboration, approval flows, handoffs, and human-agent teamwork.
Proposed solution
Add persistent relationship modeling between agents and between agents and humans, and expose that relationship context to agents when they operate.
Expected behavior:
Suggested acceptance criteria:
Willing to contribute?