An evidence-led product story
The browser
became a story.
Not another black box for agents. ThinkRun began with a simple belief: if an agent is going to change your work, you should be able to see what it saw.
Scroll to replay
The first frame
What if a browser session could be evidence—not just output?
That question pulled the product from browser automation into local control, replayable recordings, structured sharing, and an agent that can finally be shown the thing you mean.
The receipt
Built in public, measured from the work.
Every number here comes from ThinkRun’s checked-in development timeline, calculated from raw history and its memory wiki as of 9 July 2026.
Six chapters
A product is a trail of decisions.
Tap a chapter. The bars are scaled to the verified commit counts for each phase, not a growth chart.
01
Aug—Nov 2025 · 38 commitsGenesis: the initial browser automation service, session management, and its first merged pull request—#1.
The turn
From “run this” to “see this.”
ThinkRun made the browser legible: first to its operator, then to a teammate, then to the agent invited to fix what the recording proves.
thinkrun.ai/s/a-browser-session-with-context
Replayable artifact
A bug report
with eyes.
Screen, voice, clicks, console, and network in one time-synced thread—not a vague ticket someone has to reconstruct.
The point of view
Trust is not a status.
It is a replay.
That’s the through-line of the build: agents should act with context, people should stay in the loop, and a moment on the screen should travel farther than a paragraph trying to describe it.