The agent that turns websites into APIs!
---
title: Agent mode
description: Fully autonomous browser interaction. The AI browses, captures, and writes the client end-to-end.
---
Agent mode hands the steering wheel to the AI. Describe the goal in plain
English; providers such as Playwright MCP or Chrome DevTools MCP route browser actions over MCP, whereas **`agent-browser`** shells upstream's CLI directly after Reverse API Engineer resolves the binary (PATH **`agent-browser`**, else **`npm install -g <pin>`**, else **`npx -y`**) and prompt-injected skill workflows.
<Demo
src="https://raw.githubusercontent.com/kalil0321/reverse-api-engineer/main/assets/rae-autoscout.gif"
alt="Agent mode capturing a job board autonomously"
caption="Agent mode: autonomous capture and codegen in one flow."
/>
## When to use it
- You want full automation, no manual browsing required.
- The site exposes a stable, predictable flow (public job boards, docs
sites, marketing APIs).
- You're scripting captures from another agent, a cron job, or CI.
## Running interactively
```bash
reverse-api-engineer
# the REPL starts in agent mode
> click on the first job listing
```
## Running non-interactively
```bash
reverse-api-engineer agent \
--prompt "capture the public jobs api" \
--url https://example.com/jobs \
--json | jq
```
The `--json` flag makes the command emit a single JSON document on stdout
(see [Scripted usage](/docs/cli/scripted-usage) for the schema and exit
codes).
Use `--json-stream` when a supervising agent wants NDJSON progress events
during the browser and generation loop:
```bash
reverse-api-engineer agent \
--prompt "capture the public jobs api" \
--url https://example.com/jobs \
--json-stream
```
For wrapper or CI validation without launching a browser, use:
```bash
reverse-api-engineer agent \
--prompt "capture the public jobs api" \
--url https://example.com/jobs \
--dry-run
```
`--dry-run` implies JSON output and checks prompt, URL, SDK config, browser
tooling, and output directory writability.
## Agent providers
Pick the provider in `/settings` then "agent provider".
<Tabs items={['auto (default)', 'chrome-mcp', 'agent-browser']}>
<Tab value="auto (default)">
**Playwright MCP** browser automation with your configured SDK. Combines
browser control and real-time reverse-engineering in a single workflow.
Use `--headless` when running on machines without a display.
</Tab>
<Tab value="chrome-mcp">
Drives your **real Chrome** browser via
[Chrome DevTools MCP](https://www.npmjs.com/package/chrome-devtools-mcp).
Useful when you need existing sessions, cookies, or auth from your daily
Chrome.
**Requirements:**
- Chrome 146+
- Node.js 20.19+
- Auto-connect enabled at `chrome://inspect/#remote-debugging`
</Tab>
<Tab value="agent-browser">
**`agent-browser`** is the standalone [Vercel CLI](https://github.com/vercel-labs/agent-browser). Reverse API Engineer publishes a banner when it **`npm install -g`'s** the pin, verifies **`--help`**, prefers the PATH shim across runs, falls back to **`npx`** only when npm fails, injects **`skills get core --full`**, **`network har …`**, `@eN` snapshots, and **`skills list` / `skills get`** when clouds come up (see **`agent_browser_notes`**). Pins live in **`agent_browser_npx_package`** / **`RAE_AGENT_BROWSER_PACKAGE`**. Shell access is Bash-first on Claude SDK runs.
</Tab>
</Tabs>
## Reasoning model
The agent's reasoning model is the same as your configured SDK model. See
[Model selection](/docs/configuration/models) to change it.
<Callout type="info" title="Prompting tips">
- **Be specific.** *"Capture the search endpoint"* beats *"scrape the site"*.
- **Provide a `--url`.** Saves the agent a navigation hop and reduces drift.
- **Use `--json`** when calling from another agent: one JSON document on
stdout, logs on stderr, stable exit codes.
</Callout>