Summary

Tactus is a browser extension enabling AI agents to perform complex, multi-step tasks via an extensible Agent Skills system. It addresses the need for programmable automation and enhanced agent capabilities within the browser, offering a fast, token-efficient solution for developers and power users.

How It Works

Tactus integrates the Agent Skills specification into a browser extension, allowing AI to leverage imported skill folders containing instructions and JavaScript scripts. These scripts execute safely after user confirmation, enabling prompt injection for specific scenarios. The extension supports multiple AI providers (Anthropic, Gemini, OpenAI-compatible), multimodal chat with image vision, and intelligent page content extraction. A key feature is its support for the Model Context Protocol (MCP), enabling dynamic discovery and integration of external tools from HTTP servers with robust authentication. The ReAct paradigm facilitates autonomous AI tool use.

Quick Start & Requirements

• Installation: Available on Chrome Web Store and Firefox Add-ons. Manual installation via GitHub Releases or loading unpacked extension files is supported.
• Build from Source: Requires Node.js. Clone the repository, run npm install, then npm run dev (development) or npm run build (production).
• Prerequisites: Node.js for building.

Highlighted Details

• Agent Skills System: First browser extension implementing the Agent Skills specification for programmable AI actions.
• Multi-Provider AI & Vision: Native support for Anthropic, Gemini, and OpenAI APIs, plus integrated image vision.
• HTTP MCP Integration: Connects to external tool servers via Model Context Protocol for dynamic tool discovery and advanced authentication (OAuth 2.1).
• Privacy-First Local Storage: All data (conversations, skills, configs) stored locally via IndexedDB.
• Intelligent Page Understanding: Extracts core content using Readability/Turndown, with options for raw extraction.

Maintenance & Community

The project welcomes issues and pull requests. No specific details on maintainers, sponsorships, or community channels are provided. A future roadmap includes CDP automation and automated skill generation.

Licensing & Compatibility

• License: Apache-2.0 License.
• Compatibility: Permissive license suitable for commercial use and integration into closed-source projects.

Limitations & Caveats

Actively under development, Tactus currently lacks direct browser automation capabilities like Chrome DevTools Protocol (CDP), though it is planned. Automated skill generation via recording operations is also a future feature.