NotASithLord/peerd

The first AI agent harness native to the browser. A browser extension that runs a full agent loop where you already work: it drives your tabs, spins up sandboxed compute (JS notebooks, WASM Linux VMs, client-side apps), and shares what it builds peer-to-peer. BYOK, no backend, no telemetry.

What it solves

peerd provides a way to run AI agents directly within the web browser rather than relying on external platforms or hosted environments. It solves the security and accessibility problem of giving agents access to a user's active web sessions, signed-in accounts, and local compute without granting them full access to the host operating system.

How it works

It is a browser extension (Chrome and Firefox) that acts as a general-purpose agent runtime. It leverages browser primitives like Workers, sandboxing, WASM/WASI, and WebRTC to create isolated environments. The system delegates tasks to separate "actors"—scoped entities that handle specific pages or compute environments—ensuring that tools and permissions are limited to the target environment. Users can connect their own model providers (cloud-based or local via localhost/WebGPU).

Who it’s for

Users who want an AI agent that can interact with their existing web apps and browser tabs while maintaining strict security boundaries and local control over their data and model choice.

Highlights

  • Browser-Native Execution: Runs entirely inside the browser, allowing the agent to drive tabs and use signed-in sessions.
  • Isolated Compute: Supports running scripts, JavaScript Notebooks, and Linux WebVMs within browser boundaries.
  • Flexible Model Support: Compatible with various hosted providers and local models (including WebGPU).
  • Defense-in-Depth Security: Uses a vault for secrets, service-worker policy gates, and dedicated worker heaps to isolate agent loops.
  • P2P Capabilities: Preview builds include browser-to-browser discovery and agent-to-agent communication via WebRTC.

Related

  • Dispatch
    peerd: A Browser-Native AI Agent Harness

    peerd is a browser extension for Chrome and Firefox that runs a full AI agent loop entirely client-side, providing sandboxed compute and peer-to-peer sharing without a backend or telemetry.

  • Project
    hydro13/tandem-browser

    A local-first Electron browser that integrates AI agents directly into the runtime, allowing them to automate real SaaS platforms and interact with authenticated web sessions.

  • Project
    RunanywhereAI/on-device-browser-agent

    A Chrome extension AI web agent that runs entirely on-device using WebGPU, allowing users to automate web tasks privately and for free without cloud APIs.

  • Project
    browser-use/browser-use

    Browser Use is an open‑source Python/TypeScript library that gives LLM agents full control over a real or cloud‑hosted web browser. It provides 31 browser actions as tools, supports custom functions, works with OpenAI, Anthropic, Google models or its own BU2 model, and offers three usage modes: a fully hosted cloud API, a CLI for existing agents, and a library you embed in your code. Ideal for automating web tasks, building AI assistants that browse, or researching LLM‑driven web navigation.

  • Project
    S1N6H/pentest-harness

    An AI agent harness designed for authorized penetration testing and bug bounty research, featuring a multi-provider LLM engine and a built-in security toolset.