AutomatiQ MCP

An MCP server that records your browsing and network traffic so an agent can build a browserless HTTP script from the recording.

  • 179 ★ on the repository
  • Last commit
  • MIT
MCP server

Broad access

Why this level

  • Recordings keep passwords, cookies and request bodies unredacted.
  • The tool helps automate sites and collect data, which may violate the terms of a third-party service.
All reasons and checks

stonesteel27/automatiq

Install

Manual install

pip install automatiq

Python 3.11 or newer; then register the `automatiq` MCP server in the client.

Access and checks

How we review

Broad accessThe tool can change data, run commands or handle secrets.

  • Recordings keep passwords, cookies and request bodies unredacted.
  • The tool helps automate sites and collect data, which may violate the terms of a third-party service.

This is third-party code. Review the repository files before installing.

What it does

AutomatiQ reduces agent hallucinations when working with websites. You browse the target site as usual while AutomatiQ records your actions and network logs into an artifact folder. That folder becomes the source of truth: the agent writes automations and scrapers from it more accurately and faster, without spending tokens on guesses. It is a three-step process. Record: a visible Brave window opens with CDP instrumentation and capture runs until you close the last window or call `stop_recording`. Compile: network traffic is decoded into a workspace dump and, with a vision model configured, action clips get annotations. Consume: the session README documents each artifact, the MCP client reads it and writes the script. Finished scripts go straight over HTTP with no browser at runtime, so they are faster and lighter on memory. The server works over stdio and provides 5 tools. The project is in alpha.

Who it is for. Developers of automations and scrapers who must reproduce a flow on a site with no API and give an agent exact request data.

Good fit when

  • A site has no API and you must learn the request sequence behind an Export button.
  • A Playwright script is slow and should be replaced by direct HTTP calls.
  • The agent invents request parameters and needs a recording of a real session.
  • The team needs a record of the authorization flow a site actually uses.

Not a fit when

  • The site forbids automation in its terms and you use someone else account.
  • The recording would land in a shared repository: it contains passwords and cookies.
  • You need a stable production system: the project is in alpha.

Example request

Record my session on the reporting site, then use the session folder to write a Python script that exports the report through direct HTTP requests.

MCP

Transport
stdio
Authentication
not required

Limitations

Recordings are unredacted: request and response bodies, cookies and credentials, including typed passwords, are stored verbatim. Treat the session folder as a secret and never commit it. Python 3.11 or newer is required and a managed Brave downloads on first run. The project is alpha and behavior and interface change. Clip annotations need a configured vision model. The README advertises sponsor proxy services that are not needed for the tool to work. Check automation of third-party sites against terms of use and law.

How to disable. Run `pip uninstall automatiq` and remove the `automatiq` entry from the MCP client config. Delete session folders with recordings by hand: they hold passwords and cookies.

FAQ

How do I connect AutomatiQ?

Install `pip install automatiq` and add an `automatiq` server with the command `automatiq` to the MCP config. For Codex the entry goes in `~/.codex/config.toml`, for OpenCode in `opencode.json`.

How do I stop a recording?

Close the last browser window or call the `stop_recording` tool.

What is in the session folder?

Decoded network traffic, clip annotations when a vision model is configured and a `README.md` that documents every artifact for the agent.

Do I need a browser on the machine?

No, a managed Brave downloads on first run if none is found. Recording runs in a visible window with CDP instrumentation.

Is it safe to share a recording?

No. A recording holds real sessions, so the folder is neither committed nor handed to anyone.

Discussion

No questions or reviews yet.

MCP gives an agent a Playwright browser where a person can take over the same session and save an authentication profile

MCP serverBroad access927Repository stars

A CLI and MCP server that lets an agent drive your real Chrome with its logged-in sessions: 103 commands across 36 platforms plus browser automation

CLIBroad access6.3KRepository stars
Official

Browserless.io MCP server: a cloud browser for the agent with page reads, site crawls, Lighthouse audits and click-driven sessions

MCP serverBroad access19Repository stars
Official

A remote MCP server and Chrome extension: the agent reads data behind your logins and operates the sites you already use

MCP serverBroad access1.6KRepository stars

An MCP server that drives your already open Chrome and stops to ask you about a 2FA code, a CAPTCHA, or a choice only you can make

MCP serverBroad access62Repository stars
Official

Official Apify MCP server: the agent runs ready-made Apify Store scrapers for social media, maps, search and shops

MCP serverExtended access16.1KRepository stars

README in short

The project states its goal: easier reverse engineering of websites for agents. It lays out three steps, installation, MCP setup for Claude, Codex, OpenCode and oh-my-pi and a warning about sensitive data. The project vision lives in `VISION.md`.

README badge

Report a problem