Linear and Atlassian Keep the Seat and Meter What Agents Do
Both vendors keep the seat and bill agent work on top of it. Atlassian's usage billing, with pay-as-you-go on by default, starts December 3 for most meters.

TL;DR
- Linear COO Cristina Cordova says agents now create more than half of the work in Linear's issue tracker.
- Seat-based forecasts miss the spend that Linear and Atlassian now meter on top of the seat.
- Atlassian says billing for most of its usage meters, including Rovo credits, begins December 3, 2026.
Linear, which makes issue-tracking software for engineering teams, says agents now create more than half of the work in its product.
Linear COO Cristina Cordova gave the figure on a November 19 Latent Space episode and said agents created about 3% of Linear's work a year earlier. Atlassian, whose Jira competes with Linear, says usage-based billing begins December 3 for most meters. Both vendors kept the seat and added a meter for the work agents do, and a forecast built on seat count won't see the meter.
Agents and MCP clients file nearly half of issues
Linear's data page counts issues by source. In the week of August 3, 2026, agents and MCP clients created about 2,435,000 issues. People and integrations created about 2,481,000, excluding imports. Linear counts an issue as an agent's when someone delegates it to an agent or starts an agent session.
Cordova's "more than half" measures work, while the data page's near-half measures issues, and its agent column includes MCP clients. The page also says AI usage arrived as a new layer of work on top of the time people already spend in Linear. That layer is what the seat doesn't price.
On the same episode, Cordova gave Linear's company-stated figures:
- ARR above $100 million
- More than 40,000 paying customers
- 177% net revenue retention
Cordova also said Linear sells with no SDRs and no forward-deployed engineers.
Linear and Atlassian bill agent work on credits and meters
Linear was entirely seat-based until June 2026, Cordova said. Its coding sessions, launched June 11 on Basic, Business and Enterprise plans, run on AI credits. By Cordova's account on the episode, the model works like this:
- Seats stay.
- Linear Agent chat is free.
- Coding sessions and Loops run on AI credits.
Atlassian's page lists five meters as of September 1, 2026:
- Rovo credits
- Automation steps
- Assets objects
- Bitbucket meters
- Customer service AI agent resolutions
Atlassian pools the allowances at the org level and turns pay-as-you-go on by default. The seat subscription and the renewal stay as they are. A renewal signed on seats therefore sets the base fee, and the meter adds whatever the org uses above the pooled allowance.
Salesforce started charging agents to read CRM data, and Databricks gets more queries from agents than people on a meter with no seats at all. Linear and Atlassian sit between the two. The seat still carries the contract, and the agent work lands on a separate line.
Linear's own numbers say seats still sell
Cordova posted on October 6 that about 84% of Linear's weekly active users still use the app. The post also said paid seat utilization is higher than two years earlier and that weekly MCP usage is up about 20 times. Atlassian says most customers won't exceed their allowance.
The agent count is also softer than it looks. MCP clients can be driven by people, and replies to Cordova's post describe people filing issues through Claude Code and Slack chat. Some of the "agent" share is likely human work routed through an MCP client.
Linear's chart shows agent work arriving on top of human use, so rising seat use says nothing about the spend that grows outside the seat.
Linear's agent work bills on a line that seat counts don't set
Linear says coding sessions already resolve roughly 30% of its own incoming bug reports, most of them on the first pass. Each session draws on AI credits.
Linear was entirely seat-based until June 2026, Cordova said.
Atlassian starts usage-based billing for most of its new meters on December 3, with pay-as-you-go on by default.



