Blog
Notes on building and running observable browser agents — from the BrowserPilot team.
Featured · May 2026
Why we stream every agent session live
Automation people can't see is automation they don't trust. Here's why live sessions are core to BrowserPilot.
Read post →All posts
Apr 2026
Designing signed webhooks that survive a flaky endpoint
Delivery is not done when you send the request. How BrowserPilot signs, retries, and fans out run events.
Mar 2026
Per-step cost tracking, and why estimates are not enough
An agent's cost is decided step by step. BrowserPilot prices it that way instead of guessing at the end.
Mar 2026
Checkpoints and recovery: retrying without starting over
A flaky step should not cost you the whole run. How BrowserPilot resumes from the last good state.
Feb 2026
Structured output contracts beat scraping glue
Define the shape you want once, and let the agent fill it. Why typed output contracts age better than selectors.
Feb 2026
One browser per run: why isolation matters
Shared browser state is a source of flaky, hard-to-reproduce failures. Every BrowserPilot run gets its own session.
Jan 2026
From a one-off run to a scheduled task
The tasks worth automating are the ones you run again. How templates and schedules turn a run into a routine.
Jan 2026
Read-only share links: showing work without handing over a login
Sometimes a teammate or client just needs to see what happened. Share links make that a URL, not an account.
Dec 2025
Measuring agent reliability on your own tasks
A leaderboard score does not tell you whether your workflow works. Reliability has to be measured on your tasks.
Dec 2025
What a good API for browser agents looks like
Creating a run should be one POST. Observing it should not require polling guesswork. Notes on the BrowserPilot API.
Nov 2025
Browser agents vs. traditional scrapers
Scrapers encode the page; agents encode the intent. When each one is the right tool, and why intent ages better.
Nov 2025
Handling authentication in browser agents
Logins, MFA, and session reuse are where automation gets fragile. How to keep credentials scoped and runs reproducible.
Oct 2025
When to use a browser agent instead of an API
If a clean API exists, use it. The interesting question is what to do for the long tail of sites that do not have one.
Oct 2025
Running browser agents in CI
Gate a release on a real user journey. How to kick off runs from your pipeline and act on the result.
Sep 2025
Idempotent webhook handlers: a short checklist
Retries mean your handler will see the same event twice. A few habits that make duplicate deliveries harmless.
Sep 2025
What to log when an agent run fails
A failed run should answer 'what happened' on its own. What BrowserPilot keeps, and what you should add on your side.
Aug 2025
Politeness and rate limits for browser agents
An agent that hammers a site is a bad guest and a brittle one. Pacing, concurrency, and respecting the target.
Aug 2025
Persistent browser profiles: when to reuse session state
Isolation is the default for good reasons — but some workflows genuinely need continuity. How to choose.