desktop  by agentify-sh

Orchestrate AI web sessions and CLI tools locally

Created 2 months ago
278 stars

Top 93.2% on SourcePulse

GitHubView on GitHub
Project Summary

Agentify Desktop provides a local-first control center for automating interactions with web-based AI services like ChatGPT, Claude, Gemini, and Perplexity. It targets engineers and power users who leverage these AI tools alongside MCP-compatible CLI clients, enabling them to automate complex workflows directly from their local machine without managing API keys. The primary benefit is streamlined AI task execution, parallel processing, and seamless integration of web UI capabilities into command-line workflows.

How It Works

Agentify Desktop operates by controlling real, logged-in browser sessions, either through an embedded Electron instance or by attaching to a Chrome-family browser via the Chrome DevTools Protocol (CDP). It integrates with MCP-compatible CLI tools (Codex, Claude Code, OpenCode), allowing them to interact with the browser sessions. Users can manage multiple isolated workflows concurrently using stable "tab keys." The system supports practical input/output operations, including uploading local files into prompts and saving generated images or files directly to the local filesystem for subsequent reuse.

Quick Start & Requirements

  • Primary Install/Run: Clone the repository, navigate into the directory, and run ./scripts/quickstart.sh. This script handles dependency installation, MCP server registration, and launches Agentify Desktop. Alternatively, use npm i followed by npm run start.
  • Prerequisites: Node.js 20+ (22 recommended). An MCP-capable CLI tool (Codex, Claude Code, or OpenCode) is optional but required for MCP integration.
  • Links: The primary source is the GitHub repository: https://github.com/agentify-sh/desktop.

Highlighted Details

  • Supported Sites: Includes ChatGPT, Perplexity, Claude, Google AI Studio, Gemini, and Grok.
  • Browser Backend Choice: Supports default Electron or Chrome CDP, with CDP recommended for resolving Single Sign-On (SSO) issues.
  • Artifact Handling: Enables saving generated images/files locally (agentify_save_artifacts, agentify_download_images) and reusing them in subsequent prompts via attachments.
  • Codebase Context: Allows stuffing entire repositories or folders into prompts using contextPaths, with intelligent inlining of text files and auto-attachment of binaries.
  • Context Bundles: Users can save and reuse configurations including prompts, paths, and attachments via agentify_save_bundle.
  • Governor: Includes built-in rate limiting (concurrency, queries per minute, query gaps) to prevent accidental high-rate automation, configurable via the Control Center.

Maintenance & Community

The provided README does not detail specific contributors, sponsorships, partnerships, or community channels (e.g., Discord, Slack).

Licensing & Compatibility

The specific open-source license is not explicitly stated in the README. Forks and derivatives are restricted from using Agentify branding, as detailed in TRADEMARKS.md.

Limitations & Caveats

Agentify Desktop pauses automation for manual CAPTCHA verification rather than attempting to bypass them. File upload selectors, and selectors for Perplexity and Gemini, are best-effort and may require manual overrides if website UIs change. Completion detection can occasionally misfire on very long outputs or intermittent streaming pauses. Image downloads prioritize standard <img> elements, potentially missing images rendered via non-standard UI components. The experimental orchestrator/single-chat emulator is currently hidden and unsupported.

Health Check
Last Commit

2 weeks ago

Responsiveness

Inactive

Pull Requests (30d)
3
Issues (30d)
2
Star History
69 stars in the last 30 days

Explore Similar Projects

Starred by Eric Zhu Eric Zhu(Coauthor of AutoGen; Research Scientist at Microsoft Research), Elie Bursztein Elie Bursztein(Cybersecurity Lead at Google DeepMind), and
27 more.

goose by aaif-goose

14.5%
41k
Open-source AI agent for automating complex engineering tasks
Created 1 year ago
Updated 22 hours ago
Feedback? Help us improve.