Genesis
The first browser automation service: a small, direct wager that agent work needed a place to run.
ThinkRun began with a stubborn idea: when software can act for you, it should leave behind a story you can inspect.
Most software tells you it worked. ThinkRun lets you watch the evidence arrive.
Every chapter started as a test of the same belief: an agent should see enough to act — and leave enough behind to be trusted.
Not a straight line. A sequence of systems made more observable, more local, more shareable, and more honest about the messy reality of browsers.
The first browser automation service: a small, direct wager that agent work needed a place to run.
Identity, task resilience, and a video player turned execution into something people could return to, not just trigger and forget.
The Chrome extension and local bridge brought control to the browser people already use — while Loom-style sharing made the result visible beyond the terminal.
Share surfaces, embeds, and MCP parity turned a private run into an artifact that can move between tools, people, and agents.
Web recording, the Activity Hub, and billing brought the product’s clearest proposition into focus: show your agent what happened once.
Search, connector infrastructure, and pricing clarity shaped a product that can meet agents where they already work — without pretending browsers are simple.
Counts are drawn from the repository’s corrected, verified timeline: 1,571 commits, 572 unique merged PRs, 323 days from 2025-08-20, and 62 PRD thread pages. No lifetime cost or token total is shown because the record explicitly says none exists. READ THE METHOD →
ThinkRun is an attempt to make the work inspectable: the browser session, the recording, the actions, and the handoff. The story is still being written — in public, in the artifacts.