Assign it to Blaise.

Running it yourself is a CLI and a config file. Playbooks and templates cover the common shapes, and there is a skill that lets your agent do the wiring alongside you — so the thing that configures Blaise is the same kind of thing Blaise runs.

1. Linear

   Refresh the billing runbook

   Assigned to Blaise

   You assign the ticket and close the laptop.

   An assignee or a mention in a comment is the entire interface. Nobody learns a new tool, and nobody has to know which agent will do the work.

2. Blaise

   Refresh the billing runbook

   • Docs & runbooksChosen
   • Backend fixes
   • Data questions

   Why

   Documentation work, filed in Linear.

   Blaise picks the specialist.

   It reads the request and the channel it arrived on, hands the work to the agent you configured for exactly that kind of job — its own harness, tools, skills and knowledge — and records why that one.

3. Blaise

   Refresh the billing runbook

   Queued

   Waiting for a runner

   Will start on its own

   Nothing lost

   No runner up yet? It waits.

   The specialist runs on your machines, servers or CI, on your provider account and your rates. Until one of them is up, the job keeps its place, then starts itself.

4. Claude

   $ blaise runner

   ▸ session resumed · step 3

   ⎿ rate limit reached — holding

   ▸ will continue when it clears

   Out of tokens? It sits it out.

   Your harness holds on the near side of the limit and carries on from the far side, at the step it had reached.

5. Linear

   Blaise

   Two runbooks mention this step. Should I update both?

   A real question goes to a real person.

   Blaise asks in the thread rather than guessing, and then waits — so anyone on the issue can correct an assumption or unblock it, from anywhere.

6. Linear

   Refresh the billing runbook

   Opened a pull request, and answered in the thread.

   You come back to finished work.

   The result is in the thread, over a durable record of the routing decision, the tool activity, and every hold, resume and question along the way.

• BlaiseChanges

  Ship code changes

  Assign a bug or a small feature. The agent works in your repo with the plugins and skills you already use, and comes back with a branch, a pull request and a record of what it did.

  Plug in

  Your repository and CI

• BlaiseIncidents

  Chase down production issues

  Point an agent at the observability tools you already pay for. It reads the dashboards, follows the timeline and reports what it found in the issue thread, with the evidence attached.

  Plug in

  Your logs, metrics and runbooks

• BlaiseSupport

  Answer support tickets

  Give an agent your product documentation and the tickets you have already closed, and let it draft the reply. A person still decides whether it goes out.

  Plug in

  Your docs and past resolutions

• BlaiseData

  Ask questions of your data

  Put the question in an issue in plain language. The agent writes the query, runs it with the access you granted, and posts the answer alongside the SQL it used.

  Plug in

  Your warehouse or database

• BlaiseDocs

  Keep documentation honest

  Hand over the chore nobody picks up: reconcile the changelog, refresh the stale page, update the runbook after the incident that changed it.

  Plug in

  Your repo and docs site

• If you can describe the job in a prompt and reach the tools over MCP, you can put it behind Blaise.

Questions worth asking first

Do I need to keep a machine running?

You need somewhere for the runner to live. It is a lightweight CLI, and most teams start it on a small internal server or on their CI so it is available whenever someone assigns an issue; a laptop works while you are trying Blaise out. Everything else is hosted: Blaise receives the request, picks the configured specialist, keeps the conversation and stores the session activity.

How much work is it to set up?

A CLI and a config file. Install the runner, point it at your workspace, and pick the Claude Code preset — playbooks and templates cover the common shapes, and a setup skill lets your own agent do the wiring alongside you. Start on your laptop to see it work, then move the runner to a VPS or your CI when the team starts depending on it.

Can more than one person work with the same session?

Yes, and that is most of the point. A Blaise session is a Linear thread, so anyone on the issue can add context while the agent is still running, answer a question it asked, or pick the work up after a colleague. A terminal session belongs to whoever started it; this one belongs to the issue.

What data does Blaise process or store?

The request and the conversation from your work surface, the routing decision, the activity the specialist reports back, and the results it returns. Workspace, member and integration configuration too. Your source, credentials, knowledge sources and internal networks stay on your own infrastructure, where the harness runs.

Do I need my own Claude subscription?

Yes, and that is the point. Blaise runs your preferred, unmodified harness with your own Anthropic API key or Claude subscription, so you keep your rates, credits, usage limits, provider relationship and every plugin or workflow you already have. Blaise does not resell tokens or add a margin to model usage. Use a provider plan authorized for the deployment you intend.

How do I set up a speciality?

Create an agent in your workspace, choose the Claude Code preset, and give it the prompt, tools, MCP servers and repository access for that kind of work. Route it by project or issue label. You can change any of it later without touching the runner.

What can Blaise access, and who decides?

You do, when you configure the agent. Each specialist gets its own harness, tools, skills, knowledge sources and permissions, and it runs inside the security boundaries you already operate. Blaise does not widen an agent's access on its own.

What happens when a run fails or needs approval?

Blaise says so in the same thread. Durable activity history, queued prompts, heartbeats and timeouts let work resume across disconnects and process restarts. When something needs a decision, a credential or a permission it does not have, it asks and leaves the call with you rather than guessing.

How do the Blaise fee and my provider bill relate?

They are separate. Blaise charges a per-seat fee for coordination and shared access. Your model usage is billed by your provider, directly to you, at whatever rate you have negotiated. There is no Blaise markup in between.

How does the per-seat pricing work?

The Team plan includes 5 seats, and each additional seat is $4/mo on top of the base price. A seat is a person who can sign in and invoke Blaise; you set the maximum and can adjust it at any time. You can upgrade, downgrade or cancel from your account settings, and a cancellation keeps access until the end of the billing period.