Blaise is squarely in the agent ecosystem: it is an orchestration layer that turns Linear issues into agent sessions. It does not build a foundation model or an agent framework. Instead it handles the coordination problem, connecting a work item to an agent runtime, managing context, and streaming results back. The bring-your-own-Claude-subscription model makes it a thin layer on top of Anthropic's agent tooling, which means Blaise's value is in the integration and the audit trail rather than the intelligence itself.
For people building or using agents, Blaise is a concrete example of the "agent as teammate" pattern applied to a specific workflow. It shows what it looks like when agent execution is embedded in an existing tool rather than a new one, and it treats observability and replay as requirements rather than features. The self-hosted CLI and full plugin support also make it a useful reference point for teams thinking about how to keep agent compute under their own control.
Blaise is an AI agent coordinator for software teams. The product does one thing and does it inside Linear: you assign an issue to the Blaise agent or @mention it in a comment, and it spins up an agent session, pulls context from the issue, executes the task, and posts progress back as Linear comments. No new interface to learn, no separate chat window. The issue moves forward without anyone switching tabs.
The most consequential design decision is what Blaise does not do. It does not host the model. Customers run the agent CLI on their own infrastructure using their own Anthropic API key or Claude subscription. Blaise coordinates the work; the compute and the API spend stay with the customer. The company frames this as control: your usage limits, your plugins, your workflows, nothing locked inside Blaise. For engineering teams that already have Claude Code set up with custom plugins, that is a real advantage over agents that require you to adopt a new environment.
Blaise supports the full Claude Code feature set, including every plugin. It integrates with Linear, Claude, and GitHub. The Linear integration works through webhooks: Blaise registers a webhook when you authorize it, and from then on it can assign, update, and comment on issues autonomously. Administrators configure which projects and issue types the agent is allowed to act on.
Every thought, tool call, and action the agent takes is logged. Teams can replay sessions, review decisions, and maintain visibility into what agents are doing. This matters for the same reason it matters in any autonomous system: when an agent writes code or changes an issue, someone needs to be able to reconstruct what happened and why. Blaise treats that as core infrastructure rather than an afterthought.
The pricing is straightforward. Solo is free forever for one developer with one agent configuration and unlimited agent sessions. Team is $29 per month billed annually during a launch period, down from a regular $49, and includes five seats with additional seats at $4 per month each. The launch offer locks in the discounted rate for as long as the customer stays on the plan. Enterprise is custom, with dedicated onboarding, integration assistance, and priority support.
The free tier is genuinely free, not a trial. That is a deliberate land-and-expand motion: get individual developers using Blaise on their own, then convert to Team when they want colleagues in the loop or multiple agent configurations.
Blaise is betting that the agent interface of the near future is not a new app but the tools teams already use. Linear is the wedge. Whether that bet pays off depends on how many engineering teams want agent automation tightly coupled to their issue tracker, and how many would rather use a standalone agent environment. The bring-your-own-subscription model lowers the barrier to trying it and keeps Blaise out of the model-hosting business, which is a defensible position for a small team.
An AI agent coordinator that lets engineering teams assign work to AI agents directly from Linear.