Upliftr drives your app like a real user — and when a flow breaks, it traces the failure to the exact backend line that caused it, then files the fix. Run it fully self-hosted — your keys, your data, nothing leaves your infra.
What it does
One platform: chat with an AI assistant to build your tests, AI agents drive a real browser to run them, and every failure is root-caused to its exact cause — UI, API, or the exact backend line.
Just describe what to test in plain English. The assistant reads your suites, runs, and issues, writes and edits the YAML, and kicks off runs — no selectors, no page objects, no brittle scripts.
Meet the assistant →The LLM is the brain, Playwright the hands — headless or headed on Linux, macOS, or Windows.
How it works →When the UI shifts, Upliftr re-grounds against the live DOM each run instead of failing on a stale locator.
Self-heal →A failing request's trace_id is correlated to your server logs — the backend error, the likely cause, a suggested fix.
Verified failures become root-caused tickets on the built-in board — and optionally GitHub, GitLab, or Jira. Re-runs comment, never duplicate.
Filing issues →An MCP server for Cursor and Claude Code, plus a GitHub Action, GitLab CI, and a CLI that gate every PR/MR.
IDE & MCP →How it works
Author a suite, generate one from a URL or your docs, or record a click-through.
The LLM is the brain, Playwright the hands — headless on any OS. A suite shares state across its cases.
A failure is traced to its exact cause — UI, API, or the exact backend line — root-caused, and filed (deduped) where your team works.
The differentiator
“Save profile” returned a silent 500 — nothing shown to the user.
NullPointerException in upload.py:142 — avatar saved before the user row commits.
Cause + suggested fix, deduped. The ticket was waiting for you.
Author
Describe a flow the way you'd explain it to a teammate — no selectors, no page objects. Cases run in order and share fixtures, so “sign up, then log in as that user” just works.
suite: "Signup & Login" base_url: http://localhost:8500 fixtures: email: { gen: unique_email } password: { gen: password } cases: - title: "A new user can sign up" steps: - { do: "Open the sign-up page" } - { do: "Enter ${email} and ${password}" } - { do: "Click 'Create account'" } expect: - { assert: "'Account created' is shown" } - title: "That same user can log in" steps: - { do: "Log in with ${email} / ${password}" }
Built in
An MCP server puts “test this URL” inside Cursor and Claude Code.
A GitHub Action / GitLab CI and a CLI gate every PR/MR with a JUnit report and a meaningful exit code.
The built-in board, plus GitHub, GitLab, or Jira — deduped by fingerprint.
Server-log correlation turns a UI failure into a backend diagnosis.
Pricing
Source-available — run the whole platform yourself, bring your own key.
For any noncommercial use. The full platform — authoring, backend root-cause, plans, CI gating — bring your own LLM key.
Get started →For business & production use. A commercial license with terms that fit your team, plus deployment guidance.
Deploy it →Get started
Plain English in. A root-caused bug out — automatically. Self-hosted, bring your own key.
Get started →Read the docs →