Skip to content

The agent that turns websites into APIs!

3.3 KB
Raw Blame
---
title: Quick start
description: From a fresh install to a working API client in under five minutes.
---

<Steps>
  <Step>
    ### Launch the CLI

    ```bash
    reverse-api-engineer
    ```

    You land in a minimalist interactive prompt. Cycle between modes with
    **`Shift+Tab`**:

    - **agent**: autonomous AI browser agent (default)
    - **manual**: browser capture + AI generation (you drive)
    - **engineer**: re-process an existing capture
    - **collector**: AI-powered web data collection
  </Step>

  <Step>
    ### Run your first manual capture

    <Callout type="info">
      Manual mode needs the Playwright browser stack from the optional `[manual]`
      extra: `uv tool install "reverse-api-engineer[manual]"` (or add `[manual]`
      to your `pip install`) followed by `playwright install chromium`. Agent
      mode needs neither.
    </Callout>

    Press **`Shift+Tab`** once to switch from agent mode to manual mode, then
    type a goal: what you want from the site.

    ```bash
    > fetch all apple jobs from their careers page
    ```

    The CLI opens a Chromium window. You browse, click, scroll, search:
    whatever exercises the requests you care about. Close the window when
    done. The AI takes over: it reads the HAR file, identifies the API
    surface, and emits a typed client in your configured output language
    (Python by default).

    <Demo
      src="https://raw.githubusercontent.com/kalil0321/reverse-api-engineer/main/assets/reverse-api-engineer.gif"
      alt="Manual mode capture and generation"
      caption="Manual mode: you browse, the AI writes the client."
    />
  </Step>

  <Step>
    ### Read the output

    With the default Python output, generated scripts are saved in two places:

    <Files>
      <Folder name="~/.reverse-api/runs/scripts/{run_id}" defaultOpen>
        <File name="api_client.py" />
        <File name="example_usage.py" />
        <File name="README.md" />
      </Folder>
      <Folder name="./scripts/apple_jobs_api" defaultOpen>
        <File name="api_client.py" />
        <File name="example_usage.py" />
        <File name="README.md" />
      </Folder>
    </Files>

    The first is canonical (permanent, indexed by `run_id`). The second is a
    local copy with a readable name, dropped right next to your project.
  </Step>

  <Step>
    ### Run the generated client

    With the default output, the generated script is plain Python. Run it like
    any other:

    ```bash
    python scripts/apple_jobs_api/example_usage.py
    ```

    Or programmatically, via the CLI runner:

    ```bash
    reverse-api-engineer run <run_id> --file api_client.py \
      --json --auto-install
    ```

    The CLI runner also dispatches JavaScript, TypeScript, Go, Java, C#, PHP,
    Ruby, and C clients through their installed toolchains. See
    [Output language & sync](/docs/configuration/output).
  </Step>
</Steps>

## Next steps

<Cards>
  <Card title="Modes in depth" href="/docs/modes/manual" description="When to use Agent, Manual, Engineer, or Collector." />
  <Card title="Configure your model" href="/docs/configuration/models" description="Choose the model for your selected SDK." />
  <Card title="Scripted usage" href="/docs/cli/scripted-usage" description="Run from another agent with --json output." />
</Cards>