Browserberg guides

Task-oriented guides for Browserberg: connecting Playwright, agent tasks, profiles, the credential vault, workflows, triggers, pools and the live view.

How to use this section

Each guide here takes one job — connect a script, keep a login, schedule a run — and walks it end to end with executable requests. Start with Connect Playwright if you are bringing existing automation, or Run agent tasks if you are starting from a goal.

All pages in this section

1. 1

   Connect Playwright to a session

   Point an existing Playwright script at a Browserberg session with one line: connectUrl carries a short-lived access token and connectOverCDP does the rest.

2. 2

   Run agent tasks: the hosted loop

   Hand the agent a goal and let the hosted loop plan the steps. How task runs end, what the separate reviewer checks, and every option on the request.

3. 3

   Persistent profiles: keep the logged-in state

   Name a profileId and the browser state — cookies, storage, logged-in sessions — survives teardown and is restored into the next session that names it.

4. 4

   Credentials vault: secrets the model never sees

   Store portal logins so values are typed inside the browser on one registered site only — never read back, never in a prompt, a log or a replay.

5. 5

   Building workflows from blocks

   Compose navigation, actions, extraction and control flow into a versioned workflow: nine block types, immutable versions and labels that become variables.

6. 6

   Workflow healing: drift and proposals

   When a portal changes and a block starts failing, Browserberg measures the drift and proposes a fix — which a human adopts as a new version or rejects.

7. 7

   Schedule triggers: cron with a time zone

   Run a workflow on a cron schedule bound to an IANA zone. What happens on the DST switch days, after downtime, and how the firings log explains a quiet night.

8. 8

   Webhook triggers: fire a workflow from outside

   Let your ERP or CRM start a Browserberg workflow through a signed delivery: the exact HMAC recipe, the replay protection and the one identical 401.

9. 9

   Session pools: skip the cold start

   Keep browsers running ahead of demand and claim one in milliseconds. Why pool size is a total, and why a pool never falls back or changes shape.

10. 10

    Live view and takeover

    Watch any running session in the browser and take the controls when a human step comes up: the takeover lease, its heartbeats and handing control back.

11. 11

    Usage and billing data

    Read what your organisation consumed: browser minutes, task executions, gateway tokens and stored artifacts per day, plus open sessions and the included quota.

12. 12

    Watch a session without signing in

    Mint a login-free watch link for a running session: read-only by default, valid up to an hour, with the ticket in the URL fragment so it never reaches a log.

13. 13

    Connect claude.ai and other OAuth clients

    Add Browserberg as a remote MCP connector in claude.ai: the control plane is its own OAuth 2.1 server, the dashboard is the consent screen, and a grant is revocable.

14. 14

    The dashboard

    What each dashboard screen does: sessions, workflows, runs, credentials, pools, the audit log, usage, API keys and SSO, and which actions an owner keeps.

15. 15

    Report on a shop before you automate it

    Open a session, run a consent report on a German shop, read what was set before any choice, map where its forms lead, and keep a full-page screenshot.