Agent Readiness Report: Linear

Score: 60/100 · Level 2 (Agent-Aware) · scored across linear.app / developers.linear.app / linear.app/docs — 2026-05-06. Highest surface: developers.linear.app at 60/100. Marketing root: 30/100. Docs: 30/100.

Linear is the engineering-ergonomics canonical of the agent era. The changelog mentions MCP fourteen times this quarter. They invented Agent Interaction Guidelines — agent badges in the UI, dedicated Agent Sessions, an Agent Activity feed for humans to watch what their agents are doing inside a workspace. Half the engineers we know use Linear specifically because the MCP integration is the cleanest of any planning tool. None of that is in the score.

The score is also wrong, and the way it’s wrong is the most useful thing in this report.

What’s working

linear.app ships a real /llms.txt — text/plain, 9KB, properly structured. Every doc page enumerated as a .md URL. Discoverability lands at 20/25 across all three surfaces because of it.

The Agent Interaction Guidelines document is the most thoughtful public statement we’ve read on what an Agents-First product actually looks like in production: agents as first-class principals with their own avatars, activity stream, permission model. This is Level 3 thinking — shipped, in production, reachable from the homepage of a $1B+ SaaS.

What’s missing

The 60/100 on developers.linear.app is a false positive. The dev portal is a Next.js SPA. Every URL — /llms.txt, /AGENTS.md, /.well-known/mcp-server-card.json, /.well-known/ai-plugin.json, /openapi.json — returns the same 1.5MB HTML shell with status: 200. Our scorer counted those as “the artifact exists.” They don’t. v0.1.3 of the rubric needs a content-type check before crediting any “MCP Server Card published” signal. We’ll ship that fix this week.

No MCP Server Card on any surface. Linear runs an official MCP server — documented, advertised in the changelog, demoed at conferences. There is no /.well-known/mcp-server-card.json anywhere on linear.app. An agent reading the public surface cannot programmatically discover the install path.

No @linear/mcp on npm. npm view @linear/mcp returns 404. The first three results when an agent searches linear mcp are community wrappers (@hatcloud/linear-mcp, @mcp-devtools/linear, linear-mcp) — not the official server. An agent told to “install Linear’s MCP” picks a community fork by default. Ship and Forget: real product, no canonical handle on the registry agents actually search.

No bot-access-control rules. linear.app/robots.txt is 107 bytes — Disallow: /api/, Disallow: /cdn-cgi/, sitemap link, nothing else. No Content-Signal: directive. No per-bot blocks for GPTBot, ClaudeBot, Google-Extended, PerplexityBot. 0/15 across all three surfaces — the same blank score every Level 1 site has, jarring on a product this thoughtful about agents.

No markdown content negotiation. Accept: text/markdown returns HTML. Linear is publishing .md URLs in /llms.txt — they have the markdown source, they just aren’t serving it on the parent URL with the right header.

The top three fixes

  1. Publish /.well-known/mcp-server-card.json on linear.app and link it from the developers page hero. Linear runs the MCP server already; the file is a 30-line static JSON describing the install command, auth flow, and tool list. Worth 30 points across every surface and would land Linear at Level 3 instantly. Pair it with publishing @linear/mcp to npm under the official scope. See Interface First.

  2. Fix the SPA so well-known paths return real artifacts, not the React shell. Next.js rewrites for /llms.txt, /AGENTS.md, /openapi.json, /.well-known/*, /robots.txt, /sitemap.xml. Today the dev portal silently returns 200 OK on every probe and serves HTML. Worse than a 404 — agents and rubric authors both treat it as “exists” when it doesn’t. Closes a Lazy Wrapper signal nobody intended to ship.

  3. Ship /AGENTS.md and add Content-Signal + per-named-bot blocks to robots.txt. Linear’s Agent Interaction Guidelines is already the contract — it just lives at a URL the rubric can’t find. Cross-publish at /AGENTS.md and /.well-known/agent-rules. Add Content-Signal: ai-train=yes, search=yes, ai-input=yes and the standard six per-named-bot blocks. Closes the Agents Without Rules gap and picks up 15 points in bot-access-control.

What other companies can learn from this

Two lessons. For engineering-led companies: shipping the agent capability isn’t the same as letting agents discover it. Linear has done the hard part — the MCP server, the agent badges, the activity feed, the AIG document. The easy part — three static files at three well-known paths — is what gets the score. The Two Customers framing means both customers need their breadcrumbs, not just the human one.

For SPA-heavy dev portals: a 200 OK on every URL is not a feature. /.well-known/* exists specifically to be machine-checked. If your framework’s catch-all route returns the React shell for /llms.txt, you’ve made every agent’s discovery probe a coin flip and every rubric author’s life harder. Add the rewrites. Return 404 when there’s nothing there. Return the file when there is.

How we scored this

Three URLs probed via the live scorer at https://agentsfirst.dev/mcp on 2026-05-06: linear.app (30/100, Level 2), developers.linear.app (60/100, Level 2 — but see SPA caveat above), linear.app/docs (30/100, Level 2). Headline is the highest of the three. Raw probe data is in the report directory.

Methodology note: re-scored 2026-05-07 against rubric v0.2.0 — /AGENTS.md promoted from 10pts → 15pts (canonical contract artifact); /llms.txt demoted from 10pts → 5pts (10% adoption per SE Ranking, Google declined to support); /agents.json and /sitemap-index.xml now credited equally with their canonical equivalents. Section totals unchanged. The developers.linear.app SPA false-positive surfaced a v0.1.3 fix we’ll ship this week — credit “MCP Server Card published” only when the response is application/json and parses as the expected schema, not on status: 200 alone. Source: https://github.com/capitalthought/agentsfirst/blob/main/tools/agentsfirst-mcp/src/score.ts.

Part of Agent Readiness Reports — bi-weekly scorecards on how named products score against the Agents First framework. Comments, corrections, and “we just shipped the fix” notes welcome below.

💬 Comments

Have feedback, corrections, or "we just shipped the fix" notes? Comment below — backed by GitHub Discussions.