Project management for teams that build with Claude Code

If your team already uses Claude Code on client or product work, the agent is not the slow part any more. The slow part is everything around it: copying the ticket into a session, opening the PR, telling QA it's ready, finding the preview, writing release notes, and moving the status in a tracker nobody updates.

Tengu runs that loop from the ticket. One click sends a quest to Claude Code. It comes back as a pull request on the quest. QA gets a preview and a test run with screenshots. Merge and Ship happen from the board, and the quest goes live when production deploys.

Who this is for

• Agencies, product studios and in-house product teams that already pay for Claude and use Claude Code on real repos.
• Delivery leads tired of babysitting agents with a custom setup per repo.
• Teams that want the agent's work reviewed and shipped by people, with the evidence on the ticket.

How the tools compare

List prices, October 2026, per user (or per seat, for Tengu) per month, billed annually.

Tengu's prices are per seat, paid upfront, with no minimum. Client guests and agents never take a seat. Claude Code runs draw on your own Claude plan, and Tengu doesn't charge for them.

Where the others are better

• GitHub Copilot's coding agent lives inside GitHub, where your code already is. No new tool. If your team plans work in GitHub Issues and you only need issue-to-PR, it is the shortest path.
• Linear is fast, keyboard-first and loved by engineers, and its coding sessions take an issue to a pull request you can merge from Linear.
• Jira can model almost any process, and it is the incumbent.
• Plane is open source and can be self-hosted. Tengu can't.

We don't claim a better coding agent than any of them. Tengu uses Claude Code itself, on your own account. What we claim is the loop around the agent.

The loop, step by step

1. Run in Claude Code. The assignee, a PM, the project lead or an admin presses it on the quest. Tengu fires your Claude Code routine with the quest's title, description, subtasks and recent comments.
2. Draft PR with a plan. Claude works on its own branch and opens a draft PR into your base branch, with its plan. The PR attaches to the quest by itself.
3. The conversation stays on the quest. Claude's PR comments are copied to the quest. Your team writes back from the quest, and the message goes to the PR.
4. Status moves itself. PR ready: the quest moves to review. Merged: shipping. Deployed to production: live. Moves only go forward.
5. Preview and QA. The Vercel preview link lands on the quest. QA with Claude runs the repo's checks, tests the preview in a headless browser and posts findings with screenshots. It suggests a verdict and decides nothing.
6. People decide. Your QA person passes the quest or sends it back with a reason. Release people merge from the board, never into production.
7. Ship. A release PR from base to production, with a customer changelog, a QA checklist, migrations, environment changes, a rollback plan and a risk level. You merge it on GitHub; its quests go live when production deploys.

We call the result "first draft, QA'd, shipped by you." Agent PRs vary by repo. The loop puts every draft in front of a reviewer and makes every ship a person's decision.

Setup

• Connect GitHub by installing Tengu's GitHub App on your org.
• Create a Claude Code routine on your own Claude account, add an API trigger, and paste its URL and token into Tengu. Tengu never holds your Claude login.
• Link a repo to a list, with the branch PRs go into and the branch production deploys from.

Run in Claude Code, QA with Claude, Merge and Ship are on every plan, Free included; the review queue comes with Board. Claude Code routines are a research preview from Anthropic, so the trigger may change; when a run can't start, Tengu says why.

Beyond the loop

The same board carries what a delivery team needs around the code: a review queue, importers for ClickUp, Linear and Jira, Tengu Bot in Slack, and, on Studio, a timer per quest with timesheets, payroll and budgets by billable time or cost, without screenshots. Agents are never seats, so the bill counts only the people on the team.

What Tengu doesn't have yet

Custom fields of its own, docs, whiteboards, forms and a mobile app. Tengu is hosted only, with no self-hosted option, and Run in Claude Code works with GitHub repos only.