Make your agent's work faster to review, and help your team digest what shipped, by turning coding sessions into video logs.
Get early access> ship the bulk export, film it, put it on the PR ● Built. Filming the run before I open the pull request. ● Bash(qa film localhost:3000 --brief "New: bulk export") ⎿ film fj3k queued for localhost:3000 ⎿ ready: app.questline.ai/watch/fj3k
Record from your coding platform
Claude Code
Codex
Cursor
pi
Share where your team works
GitHub
Linear
curl -fsSL https://app.questline.ai/install.sh | sh
One command installs the qa tool in seconds. Your agent films any page the project runs, deployed or on localhost, and gets a link your workspace can open. Full instructions at app.questline.ai/install.
See every change your agent ships as a short film. QA it in minutes without leaving your editor.
Each session is a video plus the console errors and failed requests behind the page.
Run internal demos or a QA pass without a dev setup, and show your team the change you propose.
Agent-heavy work stays reviewable as the volume grows, because every change arrives as a recording the team can watch.
New: bulk export
$ qa shot app.example --json shot.png 2880x1686@2x 38 KB 1104 ms ! console.error: Failed to load resource: 404 (favicon.ico) ! http_error: 500 /api/preferences
One command takes a native-2x screenshot: the viewport, the full page, or one element. The console errors and failed requests from the load come with it.
“Coding agents ship more changes than anyone has time to review. A short film of each change is a faster, more human way to digest agentic coding sessions.”
We're onboarding a small pool of early users to orient the next iteration. Recording and rendering are free while we learn. Hosted sessions, shareable links, and publishing will be part of a paid plan.
Get early accessClaude Code first. Codex CLI, Cursor, pi, and any agent that runs shell commands can use the CLI.
The harness records through its own connection to the browser, outside the agent. The film, the input track, console, and network events are one record the agent cannot edit.
A session timeline with every input and its element, console and network events, the response payloads behind failures, action subtitles, and a plain-text summary the next agent can read.
The recorder is read-mostly by default. The agent submits nothing that creates, changes, or deletes server data unless you grant it.
With your account, private by default. Publishing to a pull request or a shared link is a separate, explicit step.
The same recorder runs autonomous QA sessions and renders marketing videos. Both are behind a flag while v1 stays small.
Agents are a new kind of user. They read text, run commands, and check their own work. We believe new software has to be built to equip them.
We'd love to hear what you think of the recorder, our first tool for them.
Raphael