Frequently Asked Questions

TARS uses Claude AI, which has strong support for virtually every mainstream language and framework — Python, JavaScript/TypeScript, Go, Rust, Java, Ruby, PHP, C/C++, and more. It works with any framework your project uses (React, Django, Rails, Spring, etc.) because it reads and understands your existing codebase rather than relying on templates.

No. TARS opens a pull request and notifies you, but never merges to your default branch automatically. You review the PR and merge it yourself. TARS is designed to be a force multiplier for developers, not a replacement for human judgment on what ships to production.

TARS workers clone your repository to perform their work, then delete the clone when finished. We never store your source code permanently. All communication is encrypted in transit over HTTPS. For maximum control, you can run self-hosted workers — your code never leaves your own infrastructure.

Open your project and type your request into the chat input bar at the bottom of the screen. Describe what you want in plain English — TARS reads your codebase and figures out the implementation. Press Enter or click the send button to submit.

The chat feed shows every task you've submitted as a message bubble. Each bubble displays a live status badge: queued, running, done, or failed. Bubbles update in real time as TARS works — no need to refresh the page.

Yes. You can send tasks one at a time, or use the bulk add feature to paste a list of tasks (one per line) and submit them all at once. Each task gets its own bubble in the chat feed. Tasks for different projects run in parallel; tasks within the same project are serialized to prevent merge conflicts.

The queue sidebar on the right side of the project view lists all pending and in-progress tasks for that project in order. It shows each task's position in the queue, its current status, and a short preview of the description. Completed and failed tasks drop off the queue and remain visible only in the chat feed.

When a task finishes successfully, its chat bubble updates to done and a View PR link appears inside it. Click that link to open the pull request on GitHub. TARS never merges automatically — you review and merge the PR yourself.

Simple tasks (bug fixes, adding a field, writing tests for an existing function) typically complete in 3–8 minutes. Complex tasks (new features spanning multiple files, performance refactors) can take 15–30 minutes. Queue wait time depends on available workers and how many tasks are pending.

If TARS encounters an error — tests fail, a build breaks, or Claude can't determine a safe implementation — the task is marked failed and you receive an email notification with the error details. You can review what was attempted, refine your task description, and resubmit.

Be specific. Include: what you want, where in the codebase it should live, any constraints, and acceptance criteria if possible.

Good: "Add a GET /api/users/me/ endpoint that returns the authenticated user's profile (id, email, company_name, created_at). Use the existing DRF serializer pattern in users/serializers.py."

Vague: "Add a user profile endpoint."

TARS will do its best with vague tasks, but specific descriptions lead to better outcomes.

Yes. TARS automatically discovers and runs your test suite after making changes. It looks for pytest, jest, npm test, go test, cargo test, and similar commands. If tests fail, TARS attempts to fix the failures before opening a PR.

Yes. Provide a GitHub personal access token with Contents: read & write and Pull requests: read & write scopes, or set up a deploy key with write access. See the Getting Started guide for details.

Managed workers run in TARS's cloud — nothing to set up, scales automatically.

Self-hosted workers (Pro/Team plans) run on your own infrastructure. Your code never leaves your network, you control the hardware, and you can run as many concurrent workers as you need.

TARS uses a flat monthly subscription with a task quota. If you exceed your quota, additional tasks are billed per task. See the billing page for your current plan and usage.

Yes. Cancel from the billing page anytime. Your subscription remains active until the end of the current billing period.