Expect

At a glance

Expect connects a coding agent to a real browser and uses your Git changes to focus its checks. Consider it for exploratory testing of a web app you are actively changing: it can turn a change description into browser actions and a report. Keep deterministic tests for behaviour you must repeatedly prove; an agent’s chosen checks are not a coverage guarantee.

Good fit

Developers who already have a running web app and a supported, authenticated coding agent. The executor gathers changed files, a diff preview and recent commits, then supplies that context to the agent alongside the target URL. That makes a focused request, such as checking an altered signup flow, more useful than an instruction to inspect the whole product.

Consider something else when

You need a browser test suite with fixed assertions, verified mobile coverage or a security audit. The project documents mobile testing as planned. Its advertised performance and security checks do not establish comprehensive detection, and a successful agent report cannot show that unvisited paths work.

Start here

After reviewing the installer and choosing a supported agent, the documented setup is npx -y expect-cli@latest init, followed by /expect in that agent. Start against a disposable local app with synthetic data and a narrow instruction. Use the documented --no-cookies option when existing logins are unnecessary. Check the report’s actual actions and reproduce consequential findings before accepting fixes.

Cost and requirements

The public npm metadata identifies expect-cli 0.1.3 and Node 18 or newer. A supported coding agent and its authentication are additional requirements; the agent’s subscription or model usage can impose separate costs. The repository uses FSL-1.1-MIT, with a restriction on competing commercial services and a future MIT grant, rather than unrestricted MIT terms today.

Limits to know

Local browser execution does not establish local-only processing: the source passes code context through the selected agent backend. It also records session analytics through PostHog, including agent and session information; the analytics implementation supports NO_TELEMETRY=1. This source review did not install Expect, extract cookies, run tests or measure defect detection. Cookie Sync addresses a separate need: reusing browser login cookies in a cloud browser context.

Related work

• Cookie Sync: move authorised login cookies into a cloud browser (context)
• Browser-agent evaluation: test individual actions as well as whole tasks (context)

Still unclear

• How reliably will its chosen checks catch regressions in your app’s critical flows?