About this project
Ghostget is an open-source tool designed to give AI agents precise, bounded access to web pages, media, and connected accounts. It functions as a capability and custody layer for any AI agent capable of running commands, allowing models to capture pages, preserve media, query encrypted snapshots, and use reviewed account capabilities without requiring direct access to cookies, arbitrary HTTP clients, or full browser automation.
Key features include:
- **Bounded Access**: Agents request named outcomes (e.g., `messaging.list`) rather than credentials. Ghostget binds these operations to specific providers, transports, and risk levels, stopping if facts drift.
- **Local Custody**: Archives and encrypted snapshots remain locally inspectable. Verified cached reads can function without contacting the provider.
- **Content-Bound Trust**: Plugins are approved based on verified content-addressed bundles, ensuring changed code requires new trust decisions.
- **Provider Catalog**: Supports 20+ services including Beeper, Bluesky, Facebook, GitHub, Gmail, Instagram, LinkedIn, Reddit, WhatsApp, X, and YouTube. Each service has specific adapters for official APIs or authenticated web interactions.
- **Capture & Archive**: Converts public URLs into durable Markdown, archives authorized media with integrity records, and searches local knowledge bases.
- **Persistent Reads**: Stores validated account-bound reads as encrypted snapshots, allowing revalidation without reopening browsers.
Installation requires Bun and supports macOS and Linux. The tool includes a native macOS control panel for managing accounts, permissions, and approvals. It offers both a CLI (`ghostget`) and a TypeScript SDK for integration into custom agent frameworks.
Comments
0 Rating appears after 10 ratings
Sign in to join the discussion.