Browser actions for your users · an implementation of OBEP

Run browser actions
for your users. Safely.

OpenErrand gives your product a safe, auditable way to run actions inside your users' own browsers — log in somewhere, fill a form, upload a file, pull a report. One SDK call; their credentials never leave their device.

✓ Built to Chrome Web Store standards·Apache-2.0·No remote code

// Run an action in your user's real browser. One call.
const run = client.run({ url, userId: "alex", playbookId: "acme.upload-report" });
for await (const s of run) render(s.phase);   // live status
const result = await run;                 // the result

An execution layer for products that act in the browser

Your app needs to do something on a site that has no API — for your user, in their session. OpenErrand is the safe way to run it. You keep the UI and the intelligence; you never hold their credentials.

Any site, no API needed

Drive the sites your users already use

Third-party portals, dashboards, internal tools, legacy web apps — anywhere a person logs in and does something by hand. If they can do it in a browser, your product can run it for them.

You keep the relationship

Their credentials stay theirs

Actions run in your user's own logged-in session. Logins are encrypted on their device and go straight to the destination site — never through your servers or ours.

Built for insurance, healthcare, fintech, and ops teams whose users live inside third-party portals with no API.

Safe by construction

The whole point is running actions for your users without anyone — your users, your security team, or you — having to over-trust the system.

Local-first by default

Logins are AES-GCM encrypted on the device and go straight to the destination site. They never touch our servers, so there's nothing for us to leak or subpoena.

Trust the protocol, not the vendor

Everything security-critical is open source under Apache-2.0. Your security team runs the conformance suite and verifies the relay's behavior themselves — no vendor trust required.

It is the user, present

Actions run in the user's real, authenticated session, so bot detection has nothing to flag. MFA and CAPTCHA are handled by the person who's right there.

Default-deny, cryptographically

Every task is fenced to specific domains and actions. Even a compromised relay can't widen a recipe or read a credential — the open extension re-verifies each step on-device.

How it works

From a manual process to one safe API call, in three steps.

Your user installs the extension

A lightweight Chrome extension, force-installable via Google Workspace; any logins live in an on-device encrypted vault.

You author a playbook

Record a flow once and it becomes a signed, permission-fenced recipe you register on the relay.

Your app calls one method

Your backend calls client.run(task); the action runs in your user's real browser and streams status back until it's done.

Pricing

The protocol is open — self-host it for free. You pay for the managed relay: uptime, dashboard, audit, and support.

Open source

Self-host

FreeApache-2.0 · run it yourself
  • The full OBEP reference relay
  • Extension, SDK, CLI, conformance suite
  • You operate uptime & storage
  • Community support
Get the code
Scale & compliance

Enterprise

Customannual contract
  • SSO/SAML · Workspace force-install
  • SLA, security review, BAA (HIPAA)
  • Self-host with support, or private managed
  • Dedicated support
Contact sales

You only pay for successful tasks. Prefer to run it yourself? OpenErrand is just a conformant relay — self-host the open reference relay and pay nothing.

Open by design

OpenErrand is the managed implementation of OBEP, the Open Browser Execution Protocol — an open specification with a reference implementation, not a black box. Because every security-critical component is open source and auditable, a security review doesn't have to end in "trust us." Run the conformance suite, read the extension source, and verify exactly what executes on the machine.

Give any browser process a safe, auditable API.

Run it in your users' real browsers without your product ever handling their credentials.