Autono MCP Server by MatiasV3B

by MatiasV3B• 7•• Scanned•Claim this listing•Request removal
Browser Automationagentic aiai agentai assistantanthropicantigravity

⚡ Autonomous AI Browser Agent & Navigation Copilot (Manifest V3) with Google Gemini, Anthropic Claude & OpenAI

Autono is an open-source mcp server by MatiasV3B, written mainly in JavaScript, with 7 GitHub stars and the MIT license. Its code was last updated Oct 30, 2026.

<div align="center"> <img src="Autono/assets/autono-logo.png" alt="Autono Logo" width="128" style="border-radius: 24px; margin-bottom: 8px;" /> <h1>Autono</h1> <h3>⚡ The AI browser agent that lives in your Chrome side panel</h3> <p> Chat with the page you are on, or let the AI <b>click, type and navigate for you</b>.<br/> Works with <b>Google Gemini · Anthropic Claude · OpenAI / ChatGPT</b> — through a local terminal <i>or</i> your own API key. </p> <p> <img src="https://img.shields.io/badge/Chrome_Extension-Manifest_V3-38bdf8?style=for-the-badge&logo=googlechrome&logoColor=white" alt="Manifest V3" /> <img src="https://img.shields.io/badge/Models-Gemini_%7C_Claude_%7C_ChatGPT-8b5cf6?style=for-the-badge" alt="Gemini, Claude and ChatGPT" /> <img src="https://img.shields.io/badge/License-MIT-10b981?style=for-the-badge" alt="MIT License" /> <img src="https://img.shields.io/badge/PRs-Welcome-f97316?style=for-the-badge" alt="PRs Welcome" /> </p> <p> 🌐 <b>English</b> · <a href="README.es.md">Español</a> </p> <p> <a href="#-quick-start-5-minutes">Quick start</a> · <a href="#-what-can-you-do-with-autono">Use cases</a> · <a href="#-step-by-step-setup">Setup</a> · <a href="#-using-autono">How to use</a> · <a href="#-troubleshooting--faq">FAQ</a> </p> </div>

[!TIP] If Autono saves you time, please give it a ⭐ on GitHub. It is the easiest way to help other people find it.

📖 Table of contents

  1. What is Autono?
  2. What can you do with Autono?
  3. Quick start (5 minutes)
  4. Step-by-step setup
  5. Using Autono
  6. Supported providers & models
  7. How it works
  8. Keyboard shortcuts
  9. Updating & reloading
  10. Troubleshooting & FAQ
  11. Security & privacy
  12. Repository structure
  13. Contributing · License

🌟 What is Autono?

Autono is a free, open-source Chrome extension (Manifest V3) that puts a powerful AI assistant in Chrome's side panel, right next to the page you are reading.

It has two modes:

ModeWhat it doesExample
💬 ChatReads the page you are on and answers questions about it. It never modifies the page."Summarize this article in 5 bullet points."
🤖 CoworkAn agent that plans a multi-step task and then does it in your browser: it clicks, types, scrolls and navigates, and tells you what it is doing. You can pause or take over at any moment."Find the cheapest direct flight to Madrid next Friday and fill in the search form."

Why people like it

  • 🔌 You choose the brain. Use Gemini, Claude or OpenAI models — switch with one click.
  • 🧰 Two ways to connect. Use a local terminal (the CLI of each provider, through a small local program called the Model Bridge) or paste your own API key and talk to the provider directly.
  • 🔒 Private by design. No Autono server in the middle. Your keys stay in the extension and your chats go only to the provider you chose (or to the Bridge on your own computer). Optional web search (off by default) sends the text you type to TinyFish, a third party, and only if you add a TinyFish key and turn it on. See PRIVACY.md.
  • 🧠 Control how hard it thinks. Low / Medium / High reasoning levels per model.
  • 🆓 Free and open source (MIT).

🎯 What can you do with Autono?

Real situations where Autono helps. Copy any of the example prompts.

📚 Students & researchers

  • "Explain this paper's method like I'm 15, then list its 3 main limitations."
  • "Extract every table on this page and give it to me as CSV."
  • Math and science answers are rendered with proper LaTeX formulas and syntax-highlighted code.

💻 Developers

  • Select code on any page (docs, Stack Overflow, GitHub) → click the floating ✨ Autono chip → "What does this do and where could it fail?"
  • "Write the complete code for this, no placeholders." Autono returns full files you can copy or download.
  • Use the DOM fragment picker to point at a page element and ask "why is this button misaligned?"

🛒 Shopping, travel & everyday errands (Cowork mode)

  • "Compare the three laptops open in my tabs and tell me which one is the best value."
  • "Open the booking page, search for these dates, and stop before payment."
  • Autono can ask for your approval with interactive approval cards before important steps.

✍️ Writers & office work

  • Highlight a paragraph → "Rewrite this to sound more professional."
  • "Summarize this long email thread and draft a polite reply."
  • Attach several tabs as context and ask questions across all of them.

🧪 Power users & automation fans

  • Slash commands such as /goal (run until done), /schedule (recurring tasks), /grill-me (it interviews you before acting), /teamwork and /learn (turn finished work into a reusable skill).
  • Connect MCP servers and create your own reusable skills.
  • Shadow mode: let your PC sleep or shut down automatically when a long task finishes.

⚡ Quick start (5 minutes)

Pick one path. You can always add the other later.

🅰️ API key path (simplest)🅱️ Local terminal path
You needAn API key from Google, Anthropic or OpenAIA signed-in CLI (Antigravity / Claude Code / Codex) + the Model Bridge
Install the Bridge?❌ No✅ Yes (one command)
CostPay-as-you-go with your keyUses your existing subscription / sign-in
Best forGetting started quicklyUsing models you already have access to in a terminal

The 3 steps (both paths):

  1. Get the extension → download this repo (git clone https://github.com/MatiasV3B/Autono.git).
  2. Load it in Chrome → chrome://extensions → Developer mode → Load unpacked → choose the Autono folder.
  3. Connect a model → open Autono with <kbd>Alt</kbd> + <kbd>Shift</kbd> + <kbd>A</kbd>, open ⚙️ Settings, and either paste an API key (path A) or install the Bridge (path B). Full instructions below 👇

🛠️ Step-by-step setup

Step 0 — What you need

  • Google Chrome 120 or newer (or any Chromium browser with Manifest V3 + Side Panel support).
  • Windows 10/11 if you will use the Model Bridge (path B). The extension itself works on any system Chrome runs on.
  • Git (only to download the project; you can also use Code → Download ZIP on GitHub).

Step 1 — Download the extension

git clone https://github.com/MatiasV3B/Autono.git

Step 2 — Load it in Chrome

  1. Open chrome://extensions in the address bar.
  2. Turn on Developer mode (switch in the top-right corner).
  3. Click Load unpacked and select the Autono folder inside the project.
  4. Click the 🧩 puzzle icon in Chrome and pin Autono to your toolbar.

[!NOTE] If Chrome shows "Could not load extension", make sure you selected the Autono folder (the one that contains manifest.json) and that you have the latest version (git pull).

Step 3A — Connect with an API key (no Bridge needed)

  1. Open Autono (<kbd>Alt</kbd> + <kbd>Shift</kbd> + <kbd>A</kbd> or click the icon) and open ⚙️ Settings.
  2. Paste the key of the provider you want:
ProviderWhere to create the key
🔷 Google GeminiGoogle AI Studio → API keys
🟠 Anthropic ClaudeAnthropic Console → API keys
🟢 OpenAI / ChatGPTOpenAI Platform → API keys
  1. In the model picker choose the provider's tab and switch its engine to API.
  2. Press Reload models (🔄) to load the real list of models your key can use. Done!

[!IMPORTANT] Treat API keys like passwords. Never share them or post screenshots that show them.

Step 3B — Connect with the local terminal (Model Bridge)

The Model Bridge is a small program that runs on your PC and lets Autono use the command-line tools you are already signed in to.

  1. Install it with one command. Open PowerShell and run:

    irm https://raw.githubusercontent.com/MatiasV3B/ModelBridge/main/install.ps1 | iex
    

    This installs everything in an isolated environment — you do not need to install Python — creates a desktop shortcut and starts the Bridge.

  2. Check it is running. Open http://127.0.0.1:8765/health in Chrome. You should see something like {"status":"online", ...}.

  3. Sign in to the provider you want to use (once):

    • Google Antigravity → the installer sets up the agy CLI; sign in from the Bridge app with Iniciar sesión.
    • Claude → install and sign in to Claude Code.
    • OpenAI / Codex → install and sign in to the Codex CLI.
  4. In Autono's model picker, choose the provider tab and keep its engine on Local Terminal. Press Reload models (🔄) to load what your Bridge offers.

[!NOTE] Models shown in the Antigravity tab under “Antigravity · Local Terminal” always run on your local terminal. Models under “Gemini API” always use your Gemini API key.


💡 Using Autono

Open it

  • Press <kbd>Alt</kbd> + <kbd>Shift</kbd> + <kbd>A</kbd> (A for Autono), or click the Autono icon in the toolbar.

Pick a model and how hard it thinks

Click the model button at the top:

  • Tabs: Antigravity · Claude · OpenAI.
  • Preview: the picker only describes the model (price and benchmarks).
  • Reasoning: the Reasoning button in the prompt box opens a slider above it; drag it to the level you want (Low / Medium / High, plus X-High and Max on Claude and OpenAI).
  • Engine: the Engine button in the top bar switches between Local Terminal and the provider's API.
  • 🔄 Reload models: fetches the current list from your Bridge and/or your API keys and keeps only the newest model of each family so the list stays short and clear.

Multi-agent 🧩

Turn on Settings → Multi-agent (or start one request with /teamwork) and the assistant can split a big request into sub-agents that work in parallel. Each sub-agent gets its own assistant prompt and runs in its own terminal on the Bridge (or as a separate API call in API mode); when it finishes, its result goes back to the main assistant, which writes the final answer. A card in the chat shows every agent's progress. It works with Antigravity, Claude and OpenAI, in Local Terminal or API mode.

In Cowork the same switch lets the agent open up to 50 agents, each in its own browser tab, working in parallel. Autono asks you before creating agents (you can turn that question off in Settings, but it warns that agents can use a lot of tokens).

Mascot 🦀

An optional add-on (Settings → Mascot): a little pixel critter that walks onto the tab you are on when a task starts, when it needs your OK and when it finishes, so you do not have to keep checking. It plays a different animation depending on what the task is doing (reading the page, thinking, writing code, debugging, searching, agents running, finished). When it finishes, the model writes a one-line message that appears in its speech bubble, so you can see what it did without going back to the tab. It works with every model, API or local terminal. Press <kbd>Alt</kbd> + <kbd>Shift</kbd> + <kbd>1</kbd> to call it at any time and click it to ask a quick /btw question while a task runs; it answers without touching that task. Once the task has finished, clicking it opens a normal message box that continues the same chat, and double-clicking it sends it away. You can change the shortcut at chrome://extensions/shortcuts.

Context wheel and estimated cost

Hover the ring next to the send button to see how the context is used (system prompt, page, attached files, conversation) with an estimated price for each part, the price of the next message and what the chat has cost so far. Prices are estimates from list prices (checked 2026-10-07 on the Anthropic, Google and OpenAI pricing pages), never your real bill. You can review and edit every price in Settings → Model prices. Local Terminal runs on your own subscription, so nothing is billed per token and the figures are the API-equivalent.

Chat mode 💬

Ask anything about the page. Autono reads the page content automatically (without changing it). Attach files, other tabs or a specific page element for extra context.

Cowork mode 🤖

Describe a goal and Autono will:

  1. Show you a plan.
  2. Inspect the page and carry out the steps (click, type, scroll, navigate).
  3. Report progress and the final result.

While it works, a glowing blue overlay shows the page is under Autono's control. You can Pause, Resume or Intervene at any time, and it can ask for your approval (approval cards) before important steps. A small mouse pointer travels to each element before Autono clicks or types, so you can follow it (show or hide it, and change its size, in Settings → Agent Behavior).

Handy shortcuts inside web pages

  • ✨ Context chip: select text on any page and click the floating chip to ask about it.
  • Right-click → “Ask Autono” on a selection or page area.
  • Attachment icon → DOM picker: click an element on the page to attach it.

More features

  • Slash commands: type / in the prompt box to see them (/goal, /schedule, /grill-me, /teamwork, /learn, …).
  • Dictation: press the sound-wave button in the prompt box, speak, and your words are typed into the box in the response language you chose. By default it uses the browser's speech recognition (audio goes to Google). The best engine is NVIDIA Parakeet (open source, English and Spanish, it detects the language by itself): it runs in your Bridge, is fast and accurate, and you install it from Settings → Dictation (engine about 150 MB, model about 650 MB, one time). Without it, Autono falls back to a small Whisper Tiny model that is built in (31 MB, offline, less accurate; English and Spanish). Nothing leaves your computer with either. Sound waves show while you talk, the words heard so far appear in light italics, and when you stop a spinner runs until the text is typed into the box. Test dictation checks that it works. The first time, Chrome asks for the microphone in a tab. Web search with TinyFish is off by default; turn it on in Settings (it needs your own TinyFish key).
  • Updates: every time you open Autono it checks GitHub in the background. If something is newer, a banner says New update available with an Update button: it updates the Bridge and the extension files, restarts the Bridge, reloads Autono and keeps your chats and settings. Settings → Updates and Bridge also has Check now and a hard reset button that closes the Bridge completely, frees its port and opens it again.
  • Read aloud: the speaker button under every answer reads it with Piper, a local text-to-speech engine run by your Bridge. In Settings → Voice pick a voice for your language and download it (about 20 to 115 MB, one time); the engine itself installs with one click. You can also have every answer read automatically. Piper only speaks; it cannot transcribe.
  • Chat history with folders, search and rename. Titles are generated automatically by the model.
  • Custom skills, MCP servers, and up to 5 extra OpenAI-compatible providers (for example OpenRouter or Groq).
  • Themes, languages, typography and audio options, plus an optional web-search key (TinyFish).

🧠 Supported providers & models

ProviderLocal TerminalDirect APIModels (newest versions are loaded automatically)
🔷 Google Antigravity / Gemini✅ via the Bridge✅ Gemini API keyGemini 3.8 Flash, 3.7 Flash, 3.6 Flash, 3.1 Pro · GPT-OSS 120B · Claude Sonnet 5.5 and Opus 5.5 (local terminal only) · plus every chat-capable model of your Gemini API
🟠 Anthropic Claude✅ Claude Code via the Bridge✅ Anthropic API keyClaude Sonnet 5.5 · Opus 5.5 · Fable 5.1 · Haiku 4.5
🟢 OpenAI / Codex✅ Codex via the Bridge✅ OpenAI API keyThe latest GPT and Codex models offered to your account

[!NOTE] Model lists change quickly. Autono asks your Bridge or API for the current catalog each time you press Reload models, and hides models that cannot be used for chat (for example computer-use, robotics, transcription or image-generation variants).


🏗️ How it works

flowchart TD
    subgraph Browser ["Google Chrome (Manifest V3)"]
        Panel["Autono Side Panel\n(Chat & Cowork UI)"]
        Background["Background Service Worker\n(Session & Stream Orchestrator)"]
        Content["Content Scripts & In-Page Overlay\n(Page reading & visual highlighting)"]
        Panel <--> Background
        Background <--> Content
    end

    subgraph LocalMachine ["Your computer"]
        Bridge["Model Bridge\n(http://127.0.0.1:8765)"]
    end

    subgraph Providers ["AI providers"]
        Gemini["🔷 Google Antigravity / Gemini"]
        Claude["🟠 Anthropic Claude"]
        OpenAI["🟢 OpenAI / Codex"]
    end

    Background <-->|"💻 Local Terminal"| Bridge
    Bridge <--> Gemini
    Bridge <--> Claude
    Bridge <--> OpenAI
    Background <-->|"⚡ Direct API (your key)"| Providers

In plain words: the side panel is what you see, the background worker coordinates everything, and the content scripts read and act on the web page. To reach an AI model, Autono either goes through the Model Bridge on your own computer (Local Terminal) or calls the provider directly with your API key.


⌨️ Keyboard shortcuts

ShortcutAction
<kbd>Alt</kbd> + <kbd>Shift</kbd> + <kbd>A</kbd>Open / close the Autono side panel
<kbd>Enter</kbd>Send the message / start the goal
<kbd>Shift</kbd> + <kbd>Enter</kbd>New line in the prompt box
<kbd>Esc</kbd>Close the model picker / cancel the element picker

You can change the shortcut at any time in chrome://extensions/shortcuts.


🔄 Updating & reloading

Update the extension

cd Autono
git pull origin main

Then open chrome://extensions, click 🔄 Reload on the Autono card, close and reopen the side panel, and refresh your open tabs.

Update the Model Bridge — double-click Actualizar-AntigravityBridge.bat in the Bridge folder (or run ./update.ps1). It stops the Bridge, downloads the latest code, updates only what changed and starts it again.


🩺 Troubleshooting & FAQ

<details> <summary><b>Chrome says “Could not load extension” / manifest error</b></summary>

Select the folder that directly contains manifest.json (the Autono folder). Run git pull to get the latest version and try again.

</details> <details> <summary><b>The shortcut does nothing (or opens another app)</b></summary>

Another program may already use that key combination. Go to chrome://extensions/shortcuts and set a different shortcut for Autono. Chrome only allows combinations that use <kbd>Ctrl</kbd> or <kbd>Alt</kbd>, optionally with <kbd>Shift</kbd>.

</details> <details> <summary><b>“Could not connect to Bridge” / the model list is empty</b></summary>
  1. Open http://127.0.0.1:8765/health. If it does not load, start the Bridge from its desktop shortcut or Iniciar-AntigravityBridge.bat.
  2. In Autono ⚙️ Settings check that the Bridge URL is http://127.0.0.1:8765.
  3. Press Reload models 🔄.

Using only API keys? You do not need the Bridge — make sure the provider's engine is set to API and the key is saved.

</details> <details> <summary><b>My API models do not load</b></summary>

Check that the provider's engine is set to API, that the key is correct, and press Reload models 🔄. The toast message tells you how many models each source returned or why it failed (for example “invalid API key”).

</details> <details> <summary><b>Claude from “Antigravity” does not use my Claude API key</b></summary>

By design. Claude models listed in the Antigravity tab run on your local terminal. To use the official Claude API, open the Claude tab, pick the model there and set its engine to API.

</details> <details> <summary><b>I see “Your previous response was blocked by content safety filters”</b></summary>

That message comes from the model provider, not from Autono. Autono retries the same request once automatically. If it happens again, try again, rephrase your message, or switch to another model.

</details> <details> <summary><b>Windows blocks the Bridge installer’s Python (“Application Control policy”)</b></summary>

The installer detects this and falls back to a Python you already have installed (3.10 or newer from python.org). If you have none, install one from https://www.python.org/downloads/ and run the installer again.

</details> <details> <summary><b>Does Autono work on Mac or Linux?</b></summary>

The extension works anywhere Chrome runs, with API keys. The one-click Model Bridge installer and its desktop app are Windows-first; on other systems you can run the Bridge in headless mode (see the ModelBridge repo).

</details>

🔒 Security & privacy

We do not claim Autono is “100% secure”: no software is. What we do is audit it and say plainly what we find.

  • No middleman. Autono has no server of its own. Your chats go to the provider you chose (or to the Bridge on your own computer). The only exception is the optional web search (off by default): when you turn it on and add a TinyFish key, the text you type is sent to TinyFish.
  • Your keys. API keys are stored in the extension’s local storage and are only sent to the provider you choose or to your own local Bridge. Never share them.
  • Local Bridge. It listens on 127.0.0.1 only (not reachable from other machines) and runs in an isolated uv environment, so it does not touch your system Python.
  • Dependency audit. Run cd Autono && npm audit at any time (currently 0 vulnerabilities). The source-map-js advisory (event-loop denial of service through indexed source-map offsets) was fixed by updating to source-map-js@1.2.2, and the rest of the notices came from the Tailwind CSS 3 build toolchain, which was migrated to Tailwind CSS 4. Those were devDependencies used only while styling; node scripts/build-dist.js never packages node_modules, so they were never loaded by the extension in Chrome.
  • You stay in control. Cowork shows its plan, can ask for your approval on important steps and can be paused or stopped at any time.
  • Permission modes (Settings → Permissions). They are set in the extension and work with any provider, with or without the Local Bridge; the Bridge only receives the mode when a provider runs through it. Ask manually is the default: Autono asks before it acts, and the Bridge refuses tools that need permission (opening links, deleting files). It is not a read-only mode: safe actions inside the Antigravity sandbox, such as writing a temporary file, still run without asking. Auto approve never asks, but still refuses to open local or internal addresses. Without permissions never asks and does not block local addresses. In both, Autono still never types into password, card or one-time-code fields. Actions requested by another program through remote control still ask, unless you allowed them in Tools & MCP. To turn on either mode you must type a confirmation word.

    Use them at your own risk. A web page, a file or a message the agent reads can contain instructions written to steer it (prompt injection), and with the questions turned off nobody stops it. It can put things on your computer or act on your accounts. The authors of Autono are not responsible for what happens. What protects you is the limits below, not this warning.

  • Limits no mode lifts. They live in the Model Bridge, so neither the extension nor a web page can switch them off: its Antigravity CLI always runs in its sandbox, it never shuts down or hibernates the computer, and it does not run code unless you turn that on in its own config file (code_execution in ~/.antigravity_bridge/config.json). See the Bridge README for details.
  • Found a vulnerability? Please follow SECURITY.md and report it privately.

📁 Repository structure

autono/
├── Autono/                     # The Chrome extension (load this folder in Chrome)
│   ├── assets/                 # Logos & icons
│   ├── components/             # Reusable UI components
│   ├── content/                # Content script & in-page overlay
│   ├── options/                # Options page
│   ├── side-panel/             # Side panel UI, logic and styles
│   ├── scripts/                # Packaging scripts
│   ├── background.js           # Service worker: sessions, streaming, routing
│   ├── manifest.json           # Extension manifest
│   └── package.json            # Metadata & dev dependencies
├── docs/
│   └── security/               # Security audit, test checklist and the Bridge security contract
├── tests/                      # Security and gate tests (`npm run test:unit` inside Autono/)
├── dist/                       # Packaged build (generated, not in git)
├── .github/                    # CI, Dependabot, issue & pull-request templates
├── CODE_OF_CONDUCT.md · CONTRIBUTING.md · SECURITY.md · PRIVACY.md · LICENSE
├── README.md                   # This file (English)
└── README.es.md                # Spanish version

🤝 Contributing

Contributions are welcome! Please read the Contributing Guidelines and the Code of Conduct.

  1. Fork the repository.
  2. Create a branch: git checkout -b feature/awesome-feature.
  3. Commit your changes: git commit -m "Add awesome feature".
  4. Push the branch: git push origin feature/awesome-feature.
  5. Open a Pull Request.

Found a bug or have an idea? Open an issue.


📄 License

Distributed under the MIT License. See LICENSE.

<div align="center"> <br/> <b>⭐ Star Autono if it helps you — it makes it easier for others to find it.</b> <br/><br/> <sub> Keywords: AI browser agent · Chrome side panel extension · autonomous web agent · browser automation · Gemini · Claude · ChatGPT · OpenAI · Anthropic · Google Antigravity · Manifest V3 · AI copilot · web scraping assistant · local LLM bridge </sub> </div>

Autono alternatives

View more →

Similar mcp servers in Browser Automation, ranked by GitHub stars. Compare licence, recent activity and scan status before you choose.

NameStarsLicenceLast updatedScan
Autono (this page)
MatiasV3B
7MITOct 30, 2026Scanned, no flagged issues
Chrome Devtools MCP
ChromeDevTools
55kApache-2.0Nov 1, 2026Scanned, no flagged issues
Playwright MCP
microsoft
39kApache-2.0Oct 31, 2026Scanned, no flagged issues
MCP Chrome
hangwin
13kMITJan 6, 2026Scanned, no flagged issues
Browser Tools MCP
AgentDeskAI
7.3kMITOct 28, 2026Scanned, no flagged issues
MCP
BrowserMCP
7.3kApache-2.0Oct 30, 2026Scanned, no flagged issues