Agent Interaction Guidelines and SDK (Developer Preview)
Linear released Agent Interaction Guidelines, a framework for designing agents that collaborate naturally with humans, alongside a new Agent Interaction SDK in developer preview that allows agents to communicate status, progress, and activity within Linear's interface.
Since the launch of Linear for Agents, teams have started using agents to accomplish real work inside of Linear. We've seen developers build agents that can write requirements, scope issues and projects, perform QA, fix bugs, and build features. From those early explorations, we’ve learned a lot about what makes agent collaboration work well — and where there’s still room to improve.
Agent Interaction Guidelines
We've distilled the learnings from our agent partners into a new framework called the Agent Interaction Guidelines (AIG). These principles and practices help developers design agents that integrate more naturally into human workflows.
If you're building agents and thinking through similar challenges, we encourage you to use AIG as a reference. AIG is a living document, and we expect to continually add to it as we learn more in practice.
Agent Interaction SDK (Developer preview)
To put the AIG principles into practice, we are also releasing a developer preview of our new Agent Interaction SDK. The Agent Interaction SDK gives developers a structured way to define how agents communicate status and progress. Now, developers can define how agents should communicate with the users using our SDK, and Linear renders the appropriate UI automatically.
Here's what's new in the Agent Interaction SDK:
Agent Delegation
Assignment now reflects accountability. When a user delegates an issue to an agent, the user remains the primary assignee, while the agent is added as a contributor working on the user's behalf.
Agent Session
Agents can share a rich status indicating whether they’re actively working, waiting for input, have encountered an error, or completed their work.
Agent Activity
Agents can expose detailed activity like their reasoning steps, tool usage, prompts for clarification, and final responses.
These updates give agent developers a structured, semantic way to communicate agent behavior, while giving users realtime visibility into what agents are doing — all within Linear where teams already manage their work.
The Agent Interaction SDK is available as a developer preview, and developers can start implementing the new capabilities today. To build your own agent, check out the Developing agents doc and join the #api-agents channel in our Community Slack for support.
Linear now displays stacked pull requests in Diffs with review and CI status, updates diffs in real-time as code changes, and provides admins with AI credit usage tracking and spend limit controls to manage costs.
Linear Agent now offers adaptive routing to optimize model selection by task complexity, environment secrets support for secure dependency access, and open-weight model options including GLM. Teams can define custom project labels and statuses, assign project leads, use durable IDs for projects and initiatives, and automate workflows with new team-based triggers for onboarding and reassignment.
Loops, Linear's recurring agent workflows, now respond to more workspace activity including project, initiative, and cycle changes, can edit Linear documents and post updates to Slack. This helps teams automate coordination tasks like updating docs and keeping stakeholders informed when plans change.