AI News
All stories
Reported

Linear COO says agents now create over half of its issues

Linear COO Cristina Cordova said on a podcast that agents now create more than half of issues on the platform, and that coding sessions moved to usage-based pricing in June.

Products · · 3 min read

Agents now create more than half of the issues on Linear, the issue tracker, according to COO Cristina Cordova. In a Latent Space podcast interview, she said the share was 3% a year earlier. She said Linear tracks agent activity alongside app activity, and both are rising. She uses that comparison to judge whether a product is becoming a commodity. If agent usage climbs while app usage stays flat, the product risks becoming a database behind an MCP server, a model she doesn't think is a great business.

The MCP server, the protocol-based interface that lets outside agents read and write Linear's data, shipped in May 2025. Cordova said Linear was basically the first in its category to offer one. She counts issues created in the app, meaning in the mobile, desktop and web apps, as human work. Work assigned to an agent is often counted as activity inside Linear too. That happens through a feature called loops, which automatically assigns an agent to an issue and has it start coding. In those cases, she said, the agent is a Linear agent, which is the number the company also wants to see. The podcast didn't give the share of issues that come through the MCP server alone, so the 'more than half' figure covers agents broadly rather than external tools specifically.

The most concrete change was the move away from all-seat billing. Seat fees remain, Cordova said, and until June the business was entirely seat based. Coding sessions now let users choose OpenAI or Anthropic models and harnesses. Linear passes through the published API price with a small markup, which she described as making the charges very clear and transparent. Loops are billed with AI credits. Linear Agent chat is free even though it runs on frontier models, and Cordova said the company expects to keep driving its costs down. She was candid about how hard outcome-based pricing is. Intercom's Fin, which charges for each support conversation its AI resolves, is the model everyone would love, she said. Pricing something like automatically triaging a bug is difficult, though, because companies disagree about whose job that is. So Linear mostly charges for what it costs to run. New AI features start as free betas, with notice before charges begin.

Cordova also addressed the wider market. She said AI budgets that didn't exist before are speeding up purchases, and that companies now sign one-year deals and paid pilots rather than three-year contracts. She contrasted Linear's 40,000 paying customers with companies that reach $100 million in ARR through a few contracts of $30 million to $40 million, some of them with defense companies or large law firms. The point is a caution about how that revenue is built. The podcast didn't break down how much of Linear's revenue comes from AI features, so the picture of the budgets is her view rather than something the company has quantified in the conversation.

Cordova's account of how the company sells is largely about what it avoids. Linear has never hired an SDR, the sales role that qualifies leads. She thinks the first call wastes a buyer's time with budget questions when the buyer wants a demo. Account executives do their own outbound with tooling. Linear also doesn't offer forward deployed engineering. She said the company believes its product should be easy enough to adopt without anyone getting access to a customer's systems. Solutions engineers work alongside enterprise account executives instead. Support is in-house, and she said Linear doesn't let an LLM answer tickets on its own. AI drafts replies, and a person decides what goes out. Customers sometimes assume the answers are AI-written because they arrive so quickly, she said. She claimed that in three and a half years she has never had an escalation over a poor support experience. That claim is her own and comes without any figures on ticket volume or satisfaction.

The story also has a history. Cordova said that three weeks after she joined in 2023, she crossed out an enterprise goal on a slide shown to the company. Linear had one salesperson and no realistic way to move upmarket. The company declared an enterprise push in 2025, but she said agent work displaced much of the planned enterprise roadmap. Linear built an MCP server, an agents platform, and integrations with Codex, Cursor and GitHub Copilot. She said none of these was planned at the start of 2025. She argued that racing ahead on agents would win enterprise customers in a different way than the planned features would have, because competitors were slower to build agent integrations.

What changed

Linear moved coding sessions away from all-seat billing to usage-based pricing, and users can now choose OpenAI or Anthropic models and harnesses, with seat fees still remaining.

Type: Pricing and limits · Applies to: coding sessions

Why it matters

Companies building on Linear's agent features may see a shift from flat seat fees to usage costs, and the company says new AI features may start free and later become paid.

What you can do

Check which AI features in your software contracts are free betas that may start charging, and ask vendors for notice and pricing before the trial ends.

Original source

We have no link to the original source for this story yet.

Where we saw it

1 source

Latent Space

AI wrote this summary from the sources listed, and nobody reviewed it individually before publication. Check the original source before you rely on it. How it works