weft  by jonesphillip

AI agents automate personal tasks with approval workflows

Created 3 weeks ago

New!

447 stars

Top 67.3% on SourcePulse

GitHubView on GitHub
Project Summary

Summary

Weft is a self-hosted, AI-powered task management system designed for personal use and developers. It allows users to create tasks that AI agents execute, integrating with services like Gmail, Google Docs, Sheets, and GitHub. The primary benefit is automating routine tasks while maintaining data privacy and control through deployment on Cloudflare's Developer Platform.

How It Works

Weft leverages Cloudflare's serverless architecture, comprising Workers for HTTP routing and frontend serving, Durable Objects for persistent state and real-time updates via WebSockets, and Workflows for orchestrating AI agent execution. Agents interact with integrated tools, with critical state-mutating actions (e.g., sending emails, creating PRs) requiring explicit user approval before commitment. This design ensures a private, scalable, and resilient execution environment.

Quick Start & Requirements

  • Prerequisites: Node.js 18+, Docker (for Cloudflare Sandboxes), a Cloudflare account with a paid plan (for Durable Objects/Workflows), and OAuth credentials for desired integrations (Google, GitHub).
  • Local Setup: Clone the repository, run npm install, copy .env.example to .env, and configure ENCRYPTION_KEY and OAuth secrets. Run locally with npm run dev.
  • Deployment: Deploy to Cloudflare using wrangler CLI, setting secrets via npx wrangler secret put. Authentication can be configured via AUTH_MODE in wrangler.jsonc, supporting direct use or Cloudflare Access for multi-user scenarios.
  • Documentation: Setup instructions are detailed within the README.

Highlighted Details

  • Integrations: Built-in support for Gmail (read, draft, send), Google Docs (create, edit), Google Sheets (create, update), and GitHub (issues, PRs, code).
  • Execution Environments: Utilizes Cloudflare Sandbox for isolated code execution and supports integration with remote MCP servers for custom tools.
  • Agent Workflow: Features an agent loop with automatic checkpointing and retry mechanisms.
  • User Control: All actions that mutate state require explicit user approval before execution.
  • AI Backend: Requires an Anthropic API key for agent functionality, configured per board.

Maintenance & Community

The provided README does not contain specific details regarding notable contributors, community channels (e.g., Discord, Slack), public roadmaps, sponsorships, or partnerships.

Licensing & Compatibility

Weft is licensed under the Apache License 2.0. This license is generally permissive and allows for commercial use and integration into closed-source projects, with standard attribution and license inclusion requirements.

Limitations & Caveats

Deployment necessitates a paid Cloudflare plan. Setting up OAuth credentials for Google and GitHub integrations can be a complex multi-step process. The system's core functionality relies on the availability and configuration of an Anthropic API key.

Health Check
Last Commit

2 days ago

Responsiveness

Inactive

Pull Requests (30d)
1
Issues (30d)
1
Star History
451 stars in the last 21 days

Explore Similar Projects

Starred by Chip Huyen Chip Huyen(Author of "AI Engineering", "Designing Machine Learning Systems"), Vasek Mlejnsky Vasek Mlejnsky(Cofounder of E2B), and
1 more.

pezzo by pezzolabs

0.1%
3k
Open-source LLMOps platform for streamlining AI workflows
Created 2 years ago
Updated 7 months ago
Starred by Eric Zhu Eric Zhu(Coauthor of AutoGen; Research Scientist at Microsoft Research), Mckay Wrigley Mckay Wrigley(Founder of Takeoff AI), and
24 more.

E2B by e2b-dev

0.8%
11k
Open-source cloud runtime for AI apps and agents
Created 2 years ago
Updated 3 days ago
Feedback? Help us improve.