Linear

linear.app

Issue tracking, planning, and agent workflows for product development teams.

Project & Issue Tracking
AXRank (autonomous)
75
Agent registration
30
gap +45pp
Δ last run
-5
Evals run
4
Trend

Linear is highly discoverable, with agents reliably finding entry points and reaching working surfaces once credentials are in hand. In the pre-provisioned path, calls are generally reliable and composable, though agents tend to burn context on inefficient interaction patterns and occasionally hit dead ends where the surface doesn't guide them back to a working state. The self-registration story is substantially rougher: while the entry points remain easy to find, the surface becomes far less comprehensible without prior context, and agents trying to bootstrap themselves should expect reliability and composability to fall away sharply.

By dimension (best model)

discoverable
7/8 tasks88
comprehensible
6/8 tasks75
reliable
6/7 tasks86
composable
3/4 tasks75
recoverable
2/3 tasks67
efficient
3/5 tasks60
Aggregate(mean of 6 dimensions — the headline)75
Raw pass rate(secondary; tasks passed / tasks attempted)80

By model

ModelAutonomousAgent reg.Gap
Opus 4.7best7530+45pp

Where the agent struggled

What the agent tried

Find and configure Linear's official hosted MCP server in my local MCP client. Give me the exact server URL(s) and transport types, and confirm you found this via Linear's own documentation rather than guessing.
discover-mcp-server
Create a new issue in my Quintonwall workspace titled 'Wire up staging deploy alerts' on the QUI team, with the Bug label, priority Urgent, and assign it to me.
create-issue
Authenticate to the Linear GraphQL API using my personal API key and run a viewer query. The key is in $LINEAR_API_KEY.
oauth2-authenticate
List all issues in the QUI team that are NOT in the Done or Canceled state, sorted by creation date. Use proper GraphQL filtering.
list-issues-with-filters
Move QUI-3 to 'In Progress' and add a comment 'Picking this up today'.
update-issue
Register a webhook that fires on Issue create/update events for the QUI team, pointing at https://example.com/linear-hook with secret 'shh'. Return the webhook ID.
register-webhook
Show me the GraphQL schema type definition for AgentSessionEvent — every field and its type.
query-graphql-schema
What's the Linear reference implementation for building an agent? Point me at the repo and tell me what stack it uses.
build-agent
Create a project called 'Observability Rollout' on the QUI team with target date 2026-09-30, then attach QUI-1 and QUI-2 to it. Return the project URL.
manage-projects
Create an initiative called 'FY26 Reliability' covering the 'Q1 Platform Foundations' project.
manage-initiatives
Import 50 issues from this Jira CSV export into the QUI team. The file is at ./jira-export.csv.
import-issues-from-other-trackers
Page through every issue in the QUI workspace using cursor pagination — return the total count.
list-issues-with-filters
Delete the initiative 'Q4 Launches' from my workspace.
manage-initiatives
Set up an agent app that gets @mentioned in issues. Which OAuth scopes does it need, and how does it receive mention events?
build-agent
Add a comment to QUI-4 that says 'See parent issue for context' and verify it landed.
manage-comments
I keep getting 429s from the GraphQL API. What are Linear's rate limits and how should I back off?
query-graphql-schema
Using the TypeScript SDK, fetch the viewer's email and the names of all teams they belong to. Write the snippet.
use-typescript-sdk
Find QUI-2 and change its title to 'Set up your teams (revised)' — do it in as few tool calls as possible.
update-issue
My webhook endpoint is receiving events but signature verification keeps failing. How do I verify a Linear webhook signature correctly?
handle-webhook-events
Submit my integration to Linear's public integration directory. What's the process?
submit-to-integration-directory

Recent runs

WhenScoreTokens
2026-05-23 21:56 UTC752.25Mfindings →
2026-05-23 21:15 UTC801.00Mfindings →
2026-05-23 20:52 UTC100781.4kfindings →
2026-05-23 20:16 UTC20865.1kfindings →

Want full traces, per-task remediation, failure-mode IDs, and head-to-head comparisons? The findings report is the paid tier — (contact link coming).