Building a durable research agent on Ferr
A walkthrough of an agent that researches for hours, survives restarts, and never loses its place.
Long research tasks are where browser agents either shine or fall apart. The trick is durability: the run has to survive a crash, a timeout, or a deploy without starting over.
Checkpoint everything
Store the frontier of URLs and the facts gathered so far outside the agent. When the process restarts, it reads the checkpoint and picks up where it left off instead of re-crawling.
Let Ferr hold the browser
Because the browser lives in the cloud, a restart of your agent does not kill the session. Reattach to the same session, keep the cookies, and continue the task.
- Persist the URL frontier and findings
- Reattach to a warm session after a restart
- Cap each step with an inactivity timeout
Keep reading
All articlesA beginner's guide to Ferr
From zero to your first cloud browser in a few minutes, with the code to copy.
Eight ways to build a browser agent
A tour of the patterns teams use to put a real browser behind an LLM.
Give Claude Code a real browser
Add Ferr as a tool so your coding agent can open pages, click, and read the live web.