Should Linear price AI agent usage per seat or per agent action?

Introduction

Linear charges per human seat: Free at $0, Basic at $10 per user per month, Business at $16, and Enterprise at a custom price (annual billing). That model was built for a world where people created and moved issues. In 2026 that world has changed: agents are installed in 95% of paid workspaces, roughly half of new issues are agent-created, and Linear's own products (Linear Agent, Coding Sessions, Loops, Diffs, Triage Intelligence) consume real model and compute cost every time they run. The question is whether agent usage should stay bundled into per-seat pricing or be charged per agent action.

Step 1

Clarifying Questions (3 minutes)

• Scope: Are we pricing Linear's own agent features (Linear Agent, Coding Sessions, Loops) or also third-party agents like Cursor, Devin, Codex, and Copilot that are delegated issues through the agent platform?

Why it matters: Third-party agents bring their own model cost; Linear's own agents carry Linear's inference cost. Expected answer: Mainly Linear's own agent features, with third-party agents staying free to connect. Impact on approach: The trade-off is about monetizing Linear-run compute, not taxing the ecosystem.

• Cost: What is our gross margin today on heavy agent users versus light ones?

Why it matters: If a few workspaces drive most inference cost, flat seats subsidize them. Expected answer: Heavy Coding Sessions users cost several times more than the median Business seat. Impact on approach: Some usage component is likely needed to protect margin.

• Customer: How do buyers budget for AI today? Do engineering leaders have a separate AI tools budget?

Why it matters: Usage pricing fits a variable AI budget; seats fit a headcount-based software budget. Expected answer: Both exist; large customers want predictability, AI-native startups accept usage. Impact on approach: Points toward a hybrid with spending controls.

• Strategy: Is the goal revenue capture, margin protection, or adoption of agents?

Why it matters: Per-action pricing can suppress the very behavior Linear's strategy depends on. Expected answer: Adoption first, with margin protection. Impact on approach: Avoid friction at the moment a user delegates an issue.