Agent configuration

Agent mode runs an AI loop. auto and chrome-mcp attach browser automation via MCP. agent-browser instead delegates browsing to the Vercel agent-browser CLI; Reverse API Engineer looks for agent-browser on PATH, otherwise runs npm install -g … (using the same pin as config), surfaces a console notice, verifies --help, and only falls back to npx -y … if npm cannot complete the install.

Agent providers

chrome-mcp (default)

Default for new configurations. An existing ~/.reverse-api/config.json keeps the provider it already stores.

Drives your real Chrome browser via Chrome DevTools MCP, so existing sessions, cookies, and auth from your daily Chrome profile are available. The agent opens its own tab and reads traffic through MCP tools; no HAR file is saved, so engineer <run_id> cannot regenerate these runs.

Requirements:

• Node.js 20.19+ (headed and headless)
• Headed runs only: Chrome 146+, open, with auto-connect enabled at chrome://inspect/#remote-debugging; accept Chrome's connection prompt when the run starts

Reverse API Engineer checks that Chrome's remote debugging port is reachable before launching (it does not check the Chrome or Node version) (agent --dry-run reports it as chrome-mcp:auto-connect) and stops with setup steps if Chrome is not reachable. With --headless, Chrome DevTools MCP starts its own headless Chrome instead, without your sessions.

auto

Playwright MCP browser automation with real-time reverse engineering. The configured SDK uses Playwright via MCP to control Chromium, capturing the HAR as it goes, so engineer <run_id> can regenerate a client from it later. Needs no Chrome setup. Add --headless when running without a display.

Works with the Claude SDK (default) or the OpenCode SDK. Copilot and Cursor SDK backends are also supported when their optional dependencies and credentials are installed.

agent-browser

Uses Vercel agent-browser as a standalone CLI—not a Reverse API Engineer MCP shim.

After the CLI is available, Reverse API Engineer embeds the resolved shell prefix in the user prompt (agent-browser … vs npx -y <pin> …) and validates with --help. Prompts steer the model through skills get core --full (retry with skills list / matching bundles when operators name a cloud/runtime backend via agent_browser_notes). Claude SDK runs widen the tool allow-list for Bash-heavy flows; OpenCode/Copilot/Cursor omit Reverse API Engineer browser MCP registration for agent-browser and rely on their own shell primitives.

Suggested JSON knobs:

{
  "agent_provider": "agent-browser",
  "agent_browser_npx_package": "agent-browser@0",
  "agent_browser_notes": "Optional operator guidance (AgentCore endpoints, corp proxy tweaks, sandbox IDs, …)"
}

You can alternatively drive the same knobs with RAE_AGENT_BROWSER_PACKAGE or RAE_AGENT_BROWSER_NOTES.

First-time infra still needs Chromium bits just like upstream docs describe (agent-browser install, --with-deps on trimmed Linux VMs).

How to switch providers

In the CLI:

/settings then "agent provider"

Or edit ~/.reverse-api/config.json:

{
  "agent_provider": "auto"
}

Reasoning model

The agent's reasoning model is the same as your configured SDK model. See Model selection to change it independently from the SDK.