Works from Linear issues
Assign an issue to Cyrus and it processes it automatically, streaming its progress back to Linear.
Cyrus is a strong coding agent: assign it a Linear issue and it runs Claude Code in its own worktree and opens a pull request for review. Tore writes fixes too, and also answers the customer, captures the bug and tells them when it is fixed.
Tore opens soon. Join the waitlist and we'll email you the day you can start.
Cyrus brings Claude Code into Linear, where many engineering teams already plan their work. If your work starts as Linear issues, it is a genuinely useful way to hand them to an agent.
Assign an issue to Cyrus and it processes it automatically, streaming its progress back to Linear.
Each task runs Claude Code in an isolated Git worktree, checks the work against the acceptance criteria and opens pull requests.
Bring your own Claude Code keys. Run it self-hosted on your machine, or hosted by Cyrus on some plans.
The code is open source, and Cyrus says it has completed an independent SOC 2 audit.
Pick Cyrus when
Both work in your code and open pull requests. The difference is where the work starts and who hears about it: Tore starts from the customer report or your error stream, and tells the customer when it is fixed.
| Step | Cyrus | Tore |
|---|---|---|
| Answers customers | Not part of it | Answers from your help center with the article cited, across email, chat widget and Slack. |
| Hands over to a person | Not part of it | Hands over with the full conversation when it is not confident. Anyone can take over at any moment. |
| Bug capturereplay, console, network | Not part of it | Reports from the chat widget arrive with replay, console and network attached. Bug capture |
| Investigation in your code | Runs Claude Code on Linear issues assigned to it. | Reads the repository you connect and finds the root cause, down to the file and line. |
| Fix and tests | Writes the code in an isolated worktree and checks it against the acceptance criteria. | Writes the fix and runs your tests. Your engineer can request changes in plain words. |
| Pull request for approval | Opens pull requests, ready for your review. | Preps the pull request. By default it waits for your engineer's approval. |
| Customers told it is fixed | Not part of it | Everyone who reported it is told automatically when the fix is live. |
| Model choice | Claude Code, with your own keys. | The right model for each job, from Claude, Codex, Gemini, Grok and more. You can set it per task. |
| Proactive errors | Starts from Linear issues assigned to it. | Errors from your error tracker are grouped by fingerprint and can open an investigation before anyone writes in. |
Based on Cyrus public pages (atcyrus.com), checked 2026-10-30. Products change: if we have something wrong, tell us and we will correct it. For plans and pricing, see Tore pricing.
Most teams keep support conversations in one tool and the code in another, joined by a person copying between them. A coding agent speeds up the code half, but someone still files the issue and tells the customer.
Support desk
Intercom, Zendesk and help desks like them. Great at the conversation, then the bug leaves the thread.
Tore
The conversation, the investigation and the reply live together, so nothing is retyped.
Coding agent
Cyrus and AI engineers like it. Strong at the fix, but the customer never hears from them.
The customer half
Tore keeps the conversation, the investigation and the fix in one thread. When the fix is live, everyone who reported the bug is told, on the channel they used.
If your engineers plan everything in Linear and you want Claude Code working issues in the background on your own keys and machine, Cyrus is built for exactly that. Tore connects to your tracker too, and adds the customer side.
During early access, you can also bring your own AI keys. Otherwise Tore picks the right model for each job, from Claude, Codex, Gemini, Grok and more.
By default it waits for your engineer's approval. Your engineer can request changes in plain words, and Tore revises the fix and re-runs the tests. You set how much it does on its own, per repository, in AI controls.
Plans, what is included and how the AI is billed are on the pricing page.
Tore answers your customers, captures the bug, preps the fix for your engineer to approve and tells everyone who reported it.