agentify-desktop
MCP server that enables AI tools to control local browser sessions for ChatGPT, Claude, and other AI services, supporting querying, navigation, file uploads, and artifact management.
README
Agentify Desktop
Agentify Desktop is a local control center for AI web sessions. It lets MCP-capable tools such as Codex, Claude Code, and OpenCode use the AI subscriptions you are already signed into, while keeping browser state, files, and automation on your machine.
What It Does
- Opens a local Agentify Control Center.
- Manages signed-in browser sessions for ChatGPT, Claude, Perplexity, Gemini, Google AI Studio, and Grok.
- Exposes MCP tools for querying a tab, reading a page, navigating, uploading files, saving artifacts, and reusing stable tab keys.
- Supports parallel tabs so different agents or tasks can use separate sessions.
- Packs local repo/file context into prompts when requested.
- Saves generated images/files locally so they can be reused in follow-up prompts.
Example Prompts After MCP Setup
Once Agentify Desktop is running and registered with your MCP client, you can ask for workflows like:
- “Use Agentify with key
repo-triageto ask ChatGPT for a second opinion on this bug, then compare its answer with your own analysis.” - “Open a Perplexity tab with key
research-auth-flowand research current OAuth best practices for desktop apps.” - “Send this implementation plan to Claude in a separate Agentify tab and summarize any risks it finds.”
- “Use Agentify to generate three UI concept images, save the images as artifacts, and return the local file paths.”
- “Open Grok and ChatGPT in separate Agentify tabs, ask both to review this API design, then compare the tradeoffs.”
- “Pack this repo into context, ask ChatGPT to identify risky files, and save the conversation under a stable tab key for follow-ups.”
- “Read the current ChatGPT page through Agentify and turn the conversation into actionable TODOs.”
Requirements
- Node.js 20 or newer
- An MCP-capable CLI if you want tool integration: Codex, Claude Code, or OpenCode
Supported Sites
chatgpt.comclaude.aiperplexity.aiaistudio.google.comgemini.google.comgrok.com
Preferred Install And Run
Start the desktop GUI without cloning this repo:
npx @agentify/desktop
Equivalent explicit GUI command:
npx @agentify/desktop gui
If you prefer a global install:
npm install -g @agentify/desktop
agentify-desktop
If you want the older repo-clone and local source workflow, use DEVELOPMENT_FROM_SOURCE.md.
MCP Server
Run the MCP server over stdio:
npx @agentify/desktop mcp
Show newly-created browser tabs while debugging:
npx @agentify/desktop mcp --show-tabs
With a global install:
agentify-desktop-mcp
agentify-desktop-mcp --show-tabs
Register With MCP Clients
Codex:
codex mcp add agentify-desktop -- npx -y @agentify/desktop mcp
Claude Code:
claude mcp add --transport stdio agentify-desktop -- npx -y @agentify/desktop mcp
OpenCode config example:
{
"mcp": {
"agentify-desktop": {
"type": "local",
"command": ["npx", "-y", "@agentify/desktop", "mcp"],
"enabled": true
}
}
}
Use --show-tabs at the end of the command while debugging:
codex mcp add agentify-desktop -- npx -y @agentify/desktop mcp --show-tabs
First Run
- Start the app:
npx @agentify/desktop
-
In the Control Center, create or show a ChatGPT tab.
-
Sign in to the target vendor in the browser window.
-
Register the MCP server with your CLI.
-
Ask your MCP client to use Agentify:
Use Agentify Desktop with tab key repo-triage.
Ask ChatGPT to summarize this repo in 8 bullets and list the top 3 risky areas to change first.
Return the answer and keep the tab key stable for follow-ups.
The core loop is:
- keep a real signed-in browser session open locally
- call it from an MCP client
- reuse a stable tab key across follow-up prompts
Useful MCP Tools
The MCP server registers agentify_* tools, including:
agentify_query: send a prompt to a stable tab and return the assistant response.agentify_read_page: read visible page text from a tab.agentify_navigate: navigate a tab to a URL.agentify_ensure_ready: wait for login, CAPTCHA, or UI readiness.agentify_show/agentify_hide: bring windows forward or minimize them.agentify_status: inspect tab and readiness state.agentify_tabs,agentify_tab_create,agentify_tab_close: manage tabs.agentify_save_artifacts,agentify_list_artifacts,agentify_open_artifacts_folder: manage generated files/images.agentify_save_bundle,agentify_list_bundles: save and reuse context bundles.agentify_add_watch_folder,agentify_list_watch_folders,agentify_remove_watch_folder: manage watched folders.
Artifact Workflow
Generate an image or file in a stable tab:
{
"tool": "agentify_query",
"arguments": {
"key": "ui-concepts",
"prompt": "Generate 3 clean UI concept images for a compact desktop developer tool. Keep backgrounds neutral and avoid text."
}
}
Save the generated images locally:
{
"tool": "agentify_save_artifacts",
"arguments": {
"key": "ui-concepts",
"mode": "images",
"maxImages": 3
}
}
Reattach one of the returned file paths in a follow-up:
{
"tool": "agentify_query",
"arguments": {
"key": "ui-concepts",
"prompt": "Use the attached concept image and create a more minimal variant with stronger contrast.",
"attachments": ["/absolute/path/to/concept.png"]
}
}
Codebase Context Workflow
Ask Agentify to pack local files or folders into a prompt:
{
"tool": "agentify_query",
"arguments": {
"key": "repo-review",
"prompt": "Summarize this codebase in 8 bullets and list the top 3 risky files to change first.",
"contextPaths": ["/absolute/path/to/repo"]
}
}
Control context size:
{
"tool": "agentify_query",
"arguments": {
"key": "repo-review",
"prompt": "Focus only on rendering and state management.",
"contextPaths": ["/absolute/path/to/repo"],
"maxContextChars": 120000,
"maxContextFiles": 80,
"maxContextInlineFiles": 30
}
}
The tool result includes packedContextSummary so you can see what was included, attached, or skipped.
Browser Backend
Agentify Desktop supports two browser backends:
chrome-cdp: launches or attaches to a Chrome-family browser over Chrome DevTools Protocol. This is the default and recommended backend.electron: embedded windows managed by Agentify Desktop. Use this only as an explicit fallback.
Chrome CDP is the default because SSO providers commonly block embedded Electron login:
npx @agentify/desktop
Optional Chrome CDP settings:
AGENTIFY_DESKTOP_CHROME_DEBUG_PORT=9333 npx @agentify/desktop
AGENTIFY_DESKTOP_CHROME_BIN="/Applications/Google Chrome.app/Contents/MacOS/Google Chrome" npx @agentify/desktop
You can also pass GUI flags:
npx @agentify/desktop gui --browser-backend chrome-cdp
npx @agentify/desktop gui --browser-backend electron
npx @agentify/desktop gui --chrome-debug-port 9333
Chrome CDP profile modes:
Agentify isolated profile: safest default.Existing Chrome profile: reuses your normal Chrome session. Fully quit Chrome first so the profile is not already locked.
CAPTCHA And Login Policy
Agentify Desktop does not bypass CAPTCHAs or use third-party solvers. If a verification or login challenge appears, automation pauses, brings the relevant window forward, and waits for you to complete the step manually.
If your account uses Google, Microsoft, or Apple SSO, keep auth popups enabled in the Control Center. If embedded login remains unreliable, use Chrome CDP.
Windows Notes
Use Node.js 20 or 22 on Windows. Agentify Desktop is tested against Windows in CI, including the npm CLI launcher path.
Chrome CDP is still the recommended backend on Windows because Google and Microsoft SSO can block embedded Electron login. Agentify looks for Chrome, Chromium, Brave, and Microsoft Edge in the usual install locations and on PATH.
If Chrome CDP cannot find your browser, set the executable explicitly:
$env:AGENTIFY_DESKTOP_CHROME_BIN = "C:\Program Files\Google\Chrome\Application\chrome.exe"
npx @agentify/desktop
Local Data And Privacy
Agentify Desktop is local-first:
- The local API binds to
127.0.0.1. - The local API requires a bearer token stored under
~/.agentify-desktop/. - Electron browser data is stored under
~/.agentify-desktop/electron-user-data/. - Chrome CDP profile data is stored under
~/.agentify-desktop/chrome-user-data/unless you choose an existing profile. - Artifacts, bundles, logs, and state are stored under
~/.agentify-desktop/.
Anyone with access to your machine account may be able to access local session data. Treat the machine account as the security boundary.
Environment Variables
AGENTIFY_DESKTOP_STATE_DIR: override the local state directory.AGENTIFY_DESKTOP_PORT: choose the local API port.AGENTIFY_DESKTOP_SHOW_TABS=true: show newly-created tabs by default.AGENTIFY_DESKTOP_MAX_TABS: cap parallel tabs.AGENTIFY_DESKTOP_BROWSER_BACKEND=electron|chrome-cdp: choose browser backend.AGENTIFY_DESKTOP_CHROME_BIN: choose Chrome/Chromium executable.AGENTIFY_DESKTOP_CHROME_DEBUG_PORT: choose Chrome debug port.AGENTIFY_DESKTOP_CHROME_PROFILE_MODE=isolated|existing: choose Chrome profile mode.AGENTIFY_DESKTOP_CHROME_PROFILE_NAME: choose an existing Chrome profile name.
Development From Source
Source checkout, quickstart script usage, local build commands, and source-only debugging notes live in DEVELOPMENT_FROM_SOURCE.md.
Package Commands
The npm package exposes these commands:
agentify-desktop: default GUI launcher, withmcpsubcommand support.agentify-desktop-gui: explicit GUI alias.agentify-desktop-mcp: explicit MCP alias.
Examples:
npx @agentify/desktop
npx @agentify/desktop mcp
npx -p @agentify/desktop agentify-desktop-mcp
License And Trademarks
The code is licensed under MPL-2.0. Agentify trademarks and branding are not included in that license. See TRADEMARKS.md.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.