"Linear AI" isn't a separate product. It's the AI built into Linear, the issue tracker, and in 2026 that means a built-in agent, AI triage, scheduled agent "Loops," cloud coding sessions, an official MCP server and a directory of third-party coding agents you can assign issues to. This guide explains each Linear AI feature, which plan includes it, what it costs, where it stops, and how to add a general AI agent when the work starts outside Linear.
Disclosure: This is an independent explainer published by DeskFerry. We're not affiliated with or endorsed by Linear. Linear's own AI page and docs are the official sources; everything below was checked against them in October 2026.
What is Linear AI?
Linear is a project and issue tracking tool used mostly by software teams. Over the last two years it has added AI in two directions:
- AI built by Linear, mainly Linear Agent, which launched in March 2026, plus features like Triage Intelligence that run in the background.
- A platform for other people's agents. Since Linear's "Linear for Agents" release in May 2025, third-party AI agents can join a workspace, be assigned issues and post their progress like any teammate.
When people search "Linear AI," they usually mean one of these. Both are covered below.
Linear AI features in 2026
Linear Agent
Linear Agent is the built-in assistant. You open it with Cmd/Ctrl + J, mention @Linear in an issue or document, or mention it in Slack or Microsoft Teams. According to Linear's docs, it can:
- Search issues, projects, documents and (with Code Intelligence) code
- Create and update issues, projects, milestones and initiatives
- Turn a Slack or Teams discussion into a scoped issue
- Draft project updates covering progress, risks and blockers
- Suggest what to pull into the next cycle based on backlog themes
- Save a useful conversation as a reusable "skill"
- Pull in outside context from MCP servers your admin connects, such as Notion, Granola or Figma
It works inside your existing permissions, so it can't see issues you can't. Admins can turn it off under Settings, AI & Agents.
Triage Intelligence
Triage Intelligence reads new issues as they land in Triage and suggests a team, project, assignee, labels, and related or duplicate issues, with its reasoning available on request. You can accept suggestions one by one or set each property to auto-apply per team. Linear says processing takes one to four minutes per issue and that it doesn't train its own models on your data. It's Business and Enterprise only; lighter quick suggestions are on every plan.
Loops
Loops are recurring agent runs. They trigger on a schedule (hourly to yearly), on changes to issues, projects, initiatives or cycles, or when an issue enters Triage and matches conditions. Linear's examples include investigating new bugs for a root cause, running a root-cause analysis when an incident closes, and posting weekly digests to Slack. Loops use AI credits and pause if the workspace runs out.
Coding sessions
Assign an issue to Linear Agent and, if coding sessions are enabled, it reads your codebase, writes the change in a cloud sandbox and returns a reviewable diff. Since an August 2026 update, it can also set up and run your project, test changes in a browser with before-and-after screenshots, and fix what it finds. Admins can set daily, weekly or monthly spend limits.
Code Intelligence
Code Intelligence lets Linear Agent answer questions about your codebase, such as how authentication works or who wrote a piece of logic. It needs the GitHub integration with code access, respects repository permissions by default, and is on Business and Enterprise.
Third-party agents
Linear's agent directory lists more than 30 agents, including OpenAI's Codex, Cursor, GitHub Copilot's coding agent, Devin, Factory, Sentry's agent and ChatPRD. An admin installs an agent and picks which teams it can access. After that, anyone on those teams can assign it issues or @mention it. Agents can't sign in to the app, use admin features or manage users, and Linear says they aren't counted as billable seats. The agent vendor may charge separately.
Linear MCP server
The Linear MCP server lets outside AI tools read and write Linear data. Key details:
- Endpoint:
https://mcp.linear.app/mcp(there's also a read-only endpoint) - Auth: OAuth 2.1, or an API key or OAuth token in the Authorization header
- What it can do: find, create and update issues, projects and comments
- Documented clients: Claude (claude.ai, desktop and Claude Code), Codex, Cursor, VS Code, Windsurf, Zed, v0 and Jules, plus any MCP client through
mcp-remote
MCP access is listed on every plan, including Free.
Related features: Pulse, Insights and Asks
A few features often grouped with Linear AI aren't strictly AI. Pulse is a feed of project and initiative updates with daily or weekly summaries you can listen to, on all plans. Insights is Linear's analytics and reporting, on Business. Linear Asks turns requests from Slack, email or web forms into issues, on Business, with web forms on Enterprise. Asks is an intake tool; its docs don't describe it as AI.
Linear AI pricing by plan
Linear's pricing page lists four plans. Each includes everything in the plan before it.
| Plan | Price | AI and agent features |
|---|---|---|
| Free | $0 (2 teams, 250 issues) | Linear Agent, agent platform, MCP access, Pulse, Diffs |
| Basic | $10 per user/month, billed yearly | Adds coding sessions and Loops (both use AI credits), guided reviews |
| Business | $16 per user/month, billed yearly | Adds Triage Intelligence, Code Intelligence, Insights, Linear Asks, Zendesk and Intercom integrations |
| Enterprise | Custom, annual billing | Adds SAML, SCIM, audit log, advanced Asks and other admin controls |
Plans and features on Linear's pricing page.
AI credits
Coding sessions and Loops are paid separately through AI credits, a prepaid dollar balance for the workspace. Using them is opt-in; if you never add funds, you're never charged.
- Coding sessions: model tokens at the provider's published rates with no markup, plus $0.25 per 20-minute block of sandbox runtime
- Loops: typically $0.07 to $0.20 per run without a coding session
- Top-ups: from $10, or automatic reloads from $50
- Expiry: added funds expire 12 months after purchase
AI credit pricing on Linear's docs. Every other Linear AI feature is included in the plan price.
What AI agents can do in Linear
Put together, here's what AI can realistically handle inside Linear today:
- Intake: turn Slack or Teams threads into issues with @Linear, and route new issues with Triage Intelligence.
- Triage: suggest owners, labels and projects, and flag duplicates before they pile up.
- Investigation: run a Loop on new bugs to look for a root cause.
- Implementation: hand a well-scoped issue to Linear Agent, Cursor, Codex, Copilot or Devin and review the pull request.
- Reporting: draft project updates and summarise what's at risk.
- Planning: suggest what goes into the next cycle.
The pattern that works is the same as with any AI teammate: small, clearly written issues, a person reviewing the output, and spend limits on anything metered.
Limits of Linear AI
Linear's AI is good at Linear work. A few limits are worth knowing before you plan around it:
- The best triage features are on Business. Free and Basic teams get Linear Agent but not Triage Intelligence, Code Intelligence or Asks.
- Metered work needs a budget. Coding sessions and Loops stop when credits run out, and costs vary with model usage and runtime.
- Its reach depends on what's connected. Linear Agent can pull context from MCP servers an admin connects, and Business adds Zendesk and Intercom integrations. But it's built to manage work in Linear. It isn't designed to, say, reply to the customer, update the deal in your CRM and chase the account manager once the bug ships.
- Coding agents need well-scoped issues. Vague tickets produce vague diffs. Every agent vendor, including Linear, recommends reviewing the output.
- Agents can't administer the workspace. They can't sign in to the app, change settings or manage users.
None of this is a criticism. Linear is an issue tracker, and its AI is designed to make the issue tracker better.
How to add a general AI agent to Linear for cross-tool work
A lot of what ends up in Linear starts somewhere else: a support ticket, a sales call, a Slack thread, an error alert. And a lot of what happens after an issue ships happens somewhere else too. That's where a general AI agent helps, one that works across all your tools and treats Linear as one of them.
DeskFerry is an AI agent platform with a Linear integration among its 1,500+ app integrations. You describe an agent in plain English, connect the tools it needs, and decide which actions need your approval. Some jobs teams hand off:
1. Turn customer reports into clean Linear issues. An agent reads new Zendesk or Intercom tickets, groups ones that describe the same bug, checks Linear for an existing issue, and either adds the new customer to it or files a new issue with steps to reproduce, affected accounts and links back to the tickets. For more on the support side, see AI agents for customer support.
2. Feed sales context into the roadmap. When a prospect asks for a missing feature in HubSpot, an agent logs it against the matching Linear issue so product sees the revenue behind each request. Our HubSpot to Linear page shows the connection.
3. Write the weekly status report. Every Friday, on a schedule, an agent reads what moved in Linear, adds context from GitHub and Slack, and posts a plain-English summary for people who never open Linear, such as leadership, sales or customer success.
4. Close the loop with customers. When a Linear issue moves to Done, an agent drafts a reply to every customer who reported it and holds the drafts for a person to approve through human approval before anything is sent.
5. Triage from Slack. With the DeskFerry Slack app, someone can @mention an agent on a bug thread and ask it to file the issue in Linear with the details from the thread.
How this compares with Linear Agent: use Linear Agent for work inside Linear, such as triage, planning, project updates and code. Use a general agent when the task starts or ends in another tool. Many teams will use both. For more patterns, see AI agents for project management, or browse ready-made agents in the marketplace.
How to get started
- Turn on what you already pay for. Try Linear Agent with Cmd/Ctrl + J and connect the Slack app so @Linear works in channels.
- Decide on credits. If you want coding sessions or Loops, add a small credit balance and set a spend limit.
- Pick one coding agent from Linear's directory and give it a few small, well-written issues first.
- Connect your AI tools through MCP if your team already works in Claude, Cursor or VS Code.
- Add a general agent for the cross-tool work. Start with one job, such as turning support tickets into issues, and keep approval on anything that reaches a customer.
