Tutorial

How to Run Claude Code in Parallel on Linear Issues for a Small Team (2026)

Turn a batch of Linear issues into parallel Claude Code or Codex runs, one branch and workspace per issue, and keep each issue's status and review owner clear. Paradev, Linear's own agents, and Cyrus compared.

Download Paradev

For macOS on Apple Silicon and Intel. Linux AppImage is experimental.

Last updated·

Paradev was called Superset until October 2026.

To run Claude Code in parallel on Linear issues, give each issue its own Git branch and its own agent session, start them together, and let the issue keep the status and the owner. In Paradev, open Tasks, switch the source to Linear, and click Add to workspace on an issue. Paradev opens a new workspace on Linear's branch name with the issue linked to the prompt, starts Claude Code (or Codex, or another CLI agent) on it, and moves the issue to In Progress. To start a whole batch at once, loop over the issues with the superset CLI. Each teammate runs this on their own machine, a shared host, or a Linux server in your own cloud account (Bring Your Own Cloud), with their own Claude or OpenAI subscription.

The main alternatives are Linear's own agent integrations (assign an issue to Codex or Cursor, which run in the vendor's cloud) and open-source bridges such as Cyrus, which runs Claude Code or another agent on issues assigned to it. The choice comes down to where the agent should run and who reviews the result.

This guide is written for a team of two to ten engineers who already track work in Linear. It is published by Paradev. Details about other tools come from their own docs as of October 2026.


What does a small team need from a Linear-to-agent setup?

Four things, whatever tool you pick:

  1. One branch per issue. Linear's branch name (for example eng-142-fix-export-timeout) links the branch and the pull request back to the issue automatically.
  2. One agent session per issue, so a conversation about issue 142 does not leak into issue 143.
  3. Status that moves without someone dragging cards. At minimum, In Progress when an agent starts and In Review when a pull request opens.
  4. A named human reviewer. An agent can write the change. A person on the team still owns the merge.

Git worktrees cover the first two on one machine: each issue gets its own directory and branch. They do not stop two issues from touching the same files. If issue 142 and issue 143 both edit export.ts, the second pull request to merge has a conflict to resolve, the same as with two people. Split issues by area before you hand them out.

How do I start Claude Code on a Linear issue in Paradev?

Connect Linear once under Integrations (Pro plan). Then, from the desktop app:

  1. Open Tasks in the sidebar and set the source toggle to Linear. Filter by team or assignee, for example your cycle's issues assigned to you. Keep to issues that are small and clearly described.
  2. On an issue's row, or on the issue's page, click Add to workspace.
  3. The new workspace screen opens with the issue linked in the prompt and Linear's branch name filled in. Pick the project (the repository) and the agent, such as Claude Code, add any instructions, and create the workspace.

Do this for each issue you want to hand off. Each issue gets its own workspace: a Git worktree on Linear's branch name, with its own agent session. With the integration connected, linking the issue this way moves it to In Progress.

The sidebar then shows one row per issue with the agent's state: working, waiting for input, or ready for review. Open a row to read the terminal, answer the agent, or review the diff in the Changes pane. When you are satisfied, commit, push, and open the pull request from the same pane.

How do I start a batch of Linear issues at once?

The Linear list in the desktop app works one issue at a time, so use the CLI for a batch. superset tasks get returns an issue's title, URL, branch name, and description; pass them to superset workspaces create, once per issue:

superset tasks list --tracker linear --assignee-me lists the issues to choose from. --skip-branch-prefix keeps Linear's exact branch name, so Linear links the pull request back to the issue. Use --agent codex for Codex. The script needs jq.

Can I start it from Linear itself?

Yes. From Linear, use its custom coding tool setting. A short wrapper script lets Work on issue → Custom script call superset workspaces create with the issue's branch and prompt, then superset workspaces open to show it in the desktop app. The Linear integration docs have the full script and the ~/.linear/coding-tools.json file.

Can an agent start when an issue is assigned or labeled?

Yes, with an Automation (Pro). Add a Linear trigger for an issue that is created, moves to a status, or is assigned, and scope it by team, project, label, or assignee. The status filter applies only to Issue moved to. A common pattern reads "Issue moved to Todo in Engineering with label agent": when an issue with the agent label moves into Todo, the automation creates a workspace for it on the device you picked and starts the agent with your prompt plus the issue in front of it. Adding the label alone does not fire the trigger, so add it before the issue moves.

Two limits to plan for: the target device has to be online when the event arrives, or the run is recorded as missed and is not retried. Delivery is at least once, so keep the prompt safe to run twice.

How does the team keep track of who reviews what?

Keep ownership in Linear and review in GitHub, and let Paradev connect them:

  • Linear keeps the assignee and status. With the integration connected, starting a workspace from an issue moves it to In Progress, and the branch name links the pull request back to the issue. To move issues when a pull request opens or merges, connect GitHub in Linear and set the team's pull request automations in Linear's settings.
  • Paradev's Pull requests view filters by Awaiting review from you, Changes requested, and Approved review, with each PR's checks next to its diff. Merging, closing, and commenting act on the real PR with your GitHub permissions.
  • Workspaces filter by who created them, so you can see a teammate's agent runs without mixing them with yours.

To give the agent itself tools to read and update Linear, such as posting a summary comment on the issue, install the Linear plugin.

What do other tools do for the same job?

OptionWhere the agent runsAgentsGood fit
Paradev Tasks and automationsYour machines, a shared host, or your own cloud serverClaude Code, Codex, and other CLI agentsTeams that want agents on their own machines and subscriptions, with diffs reviewed before the PR
Codex in LinearCodex cloudCodexAssign or mention @Codex on an issue, and it posts progress back
Cursor in LinearCursor Cloud AgentsCursor's agentTeams already on Cursor; assign an issue to Cursor or mention @Cursor
Linear coding sessionsA secure cloud environment, per Linear's docsLinear Agent, writing code with Claude Code or CodexTeams that want everything to stay inside Linear; needs AI credits and a workspace admin to turn it on
CyrusA machine you run, or Cyrus's cloudClaude Code, Codex, Cursor, Gemini, OpenCodeOpen-source (Apache 2.0) bridge that works issues assigned to it, one worktree per issue

Linear's native agents need the least setup and keep the conversation on the issue. The tradeoff is that the work runs in the vendor's environment, which may not have your local services, seeded database, or VPN. Paradev runs the agent CLIs you already use on machines you control, so the agent sees what your developers see, and a person reviews the diff in the app before the pull request goes up. You can use both: assign the quick, self-contained issues to a cloud agent and run the ones that need local context in Paradev.

Frequently Asked Questions

Does Claude Code integrate with Linear?

Not on its own. Claude Code reads Linear through an MCP server or a plugin. To start Claude Code from Linear issues, use a tool that bridges the two: Paradev's Tasks view, CLI, and Linear triggers, Linear's coding sessions, or Cyrus.

Can several Claude Code agents work on different Linear issues at the same time?

Yes. Give each issue its own worktree and branch so the agents do not edit each other's files. In Paradev, click Add to workspace on each issue, or run the CLI loop above for a batch. Issues that touch the same code can still conflict at merge time.

Do I need a Paradev seat for each teammate?

Each person who uses Paradev needs their own account. The Linear integration, automations, and remote hosts are on the Pro plan. See pricing.

Does Paradev move Linear issues to In Progress?

Yes, when the Linear integration is connected and you start a workspace from an issue in the app. Later moves, such as to In Review or Done, come from Linear's own GitHub integration: connect GitHub in Linear and choose which status each pull request event sets for your team.

Can I use Codex instead of Claude Code?

Yes. Pick Codex on the new workspace screen, or pass --agent codex to the CLI or the automation. Different issues can use different agents.