Published: 2026-05-17

How to Use Linear as a Second Brain for Claude Code and Codex

Chapters / key moments (click to jump — plays here on the page)

Alex Finn demonstrates a workflow that pairs Claude Code (or Codex) with Linear — a free, cloud-synced project management tool — to eliminate the context drift that plagues long vibe-coding sessions. Describe your app once in Claude Code; it auto-generates Linear projects, issues, acceptance criteria, and priorities. From then on, any device running Claude Code can say "work on the next task" and pick up exactly where the previous session left off — no re-explaining required.

Source video

"Claude Code is 1000x better when you use this tool" by Alex Finn — Watch on YouTube →

Key Takeaways

Linear is free up to a high usage ceiling and integrates natively with Claude Code and Codex via "connect apps" — no API key needed, OAuth login only.
Claude Code auto-creates Linear projects and issues (with acceptance criteria, priorities, due dates, and scope) from a plain-English app description — replaces hours of manual project setup.
Cloud-synced issue board means phone, iPad, and laptop all start from the same state — tell Claude Code "work on the next task" and it reads Linear, knows what to build, and starts immediately.
Works with both Claude Code and Codex simultaneously — both can pull tasks from the same Linear board, enabling multi-agent coordination on a single project.
Add a CLAUDE.md (or agent.md for Codex) with rules for how to use Linear — keeps behavior consistent and prevents drift across sessions.

What you can actually set up from this

Extracted from the video's own transcript — the specifics the original summary left out.

Reproducible steps

Connect Linear to the coding agent
Claude app: Customize → Connect your apps → search Linear → log in. In Codex, use the plugins menu (top left); the process is the same. Confirm the connection by asking whether it can see your Linear team.
Spec the whole app into Linear before writing code
Prompt example: 'I want to build a prompt library app (Next.js, folders, categories, optimisation). Before we begin, build out all of the issues in Linear, and projects inside the team to group them.' In the demo it created ~90 issues across prioritised projects, each with details and acceptance criteria.
Build from the board
Ask 'what should our first few issues be?', then 'take issues from backlog, work on them, and complete them'. The agent moves each issue from backlog to in progress to done and tests against its acceptance criteria. The demo completed 17–18 issues in one run.
Run Claude Code and Codex off the same board
Open the same project in both. Each reads Linear, claims the next issue and works independently, with no hand-off between agents.
One GitHub branch per issue
Each issue gets its own branch (the link is posted on the issue) and a pull request you review before merging, so parallel agents don't overwrite each other's code.
Put the rules in CLAUDE.md / AGENTS.md
Before editing, read the Linear issue and spec; stop at the acceptance criteria; don't touch unrelated files or refactor unprompted; open a pull request when done; keep the issue status updated.
Optional: Slack notifications
Separate channels for Linear issue updates and GitHub PR/merge events keep a team in sync.

Gotchas

Not sponsored, per the creator. Linear's free tier covers this usage.
The demo skipped auth and the database to stay short. Real apps need those issues too.
Without issues and acceptance criteria, the agent tends to stop after one task or drift into work you didn't ask for. The board fixes both.