Agent Interaction Guidelines (AIG) — Linear
URL: https://linear.app/developers/aig
Author: Linear
Published: 2026
Summary
Linear’s evolving principles for designing agent interactions in software development workflows. As agents produce work in abundance, the value shifts to orchestrating input, context engineering, and reviewing output. AIG defines a new contract for human-computer interaction in this world.
Key Claims
- Always disclose agent identity: Humans must have instant certainty about who they’re interacting with. Agents must never be mistakable for a person.
- Inhabit the platform natively: Agents should work through existing UI patterns and standard actions — not require new parallel interfaces.
- Instant, unobtrusive feedback: Silence creates uncertainty. Agents should immediately signal they’ve received a request.
- Full transparency of internal state: Show whether thinking, waiting, executing, or done. Humans should be able to inspect reasoning, tool calls, prompts, and decision logic on demand.
- Respect disengagement requests: When asked to stop, stop immediately. Only re-engage with a clear signal.
- Accountability stays with humans: There should be a clear delegation model. Agents carry out tasks, but final responsibility always remains with a human.
- Living document: AIG is explicitly described as evolving — Linear invites community contribution.
Why This Matters
As agents become first-class participants in engineering workflows, these principles define the UX contract. The transparency principle (inspect reasoning on demand) is particularly important — it’s what separates trustworthy agent integrations from black boxes.