open-source

A curated GitHub resource mapping AI token costs, waste, and efficiency practices for agentic coding. It combines an awesome list with editorial notes, practices, concepts, claims, and copy-paste setups.

Local-first change control for agent skills, with evidence-gated promotion, SkillOpt optimization, and MCP-based serving of skills.

Open-source family AI agent for the home, built to run on a wall-mounted Android phone. The project positions itself as a shared household assistant for shopping lists, calendars, timers, reminders, expenses, and approved family calling.

Hearth is an open-source local AI agent that can control a Windows PC: it opens apps, reads and writes files, drives a browser, uses voice, and keeps memory across sessions.

Homestead is an open-source, self-hosted platform for building and deploying personal apps for you, your family, and AI agents. The page highlights app authentication, offline support, a CLI workflow, AI features, push notifications, and bulk import, with install and docs links visible.

Figmaboy is a local-first desktop design workspace with Codex CLI built into the canvas. It combines design, review, and command-line actions in one app, with offline saved designs and MCP-based context sharing.

Open-source AI infrastructure focused on portability, governance, auditability, and measuring model cost and outcomes to reduce vendor lock-in.

Wallie is an open-source, self-hosted AI streamer that can watch and hear your screen, speak in a custom persona, and now play Minecraft autonomously while narrating its actions. The repo shows a live-streaming workflow with vision, hearing, TTS, avatars, OBS integration, and dashboard-driven setup.

A Cloudflare Worker that notifies you when visitors arrive from AI assistants like ChatGPT, Perplexity, Gemini, Claude, Copilot, or Grok, with optional alerts for AI crawlers.

Local-first task and context manager for humans and AI coding agents. It stores tasks, plans, docs, and memory in a shared SQLite database, with CLI commands, a web UI, and agent attribution built in.

A Pi extension that uses AI to generate and keep codebase documentation current, with commands for initial docs, incremental updates, and chat-based questions.

An open-source enforcement layer for AI coding agents that adds frozen specs, tamper-detected tests, independent verification, and hard-blocking gates so unverified work cannot pass.

Private skill search for AI coding agents. It helps an agent match abstract capability wishes to vetted skills without sending task details off machine.

Open-source, local-first money agent that finds waste, drafts cancellation or dispute actions, and only proceeds with user approval. It emphasizes consent, a tamper-evident audit trail, and bring-your-own-key operation.

A Pomodoro timer for Claude Code that shows a live countdown in the terminal status line and includes a backup alarm, CLI controls, and setup commands.

Open-source multi-agent protocol for AI research agents. It pairs a tool-using Scientist with a question-only advisor that can only ask questions and approve plans, and the README includes quick-start setup plus notes on reproducing results on MLE-bench/Kaggle tasks.

An open-source conversational language tutor for iOS with voice I/O, AI-powered feedback, and support for bringing your own API keys.

Buzz is a self-hostable workspace for humans and AI agents to collaborate in the same threads, with signed events, audit trails, channels, workflows, and code-oriented actions.

FERNme is an open-source memory engine for AI agents that uses fuzzy graphs and Hebbian-style updates to store user preferences with near-zero LLM calls. The repository highlights editable user-owned memory, flat token usage, outcome-based reinforcement, and deployable plumbing via REST and MCP.

Sana is an object-storage-native search database for vectors, full-text search, and attribute filters. The repo describes durability, indexing, backends, operations, and quick-start commands for running the service or demo.

A local Rust CLI for A/B testing two CLAUDE.md files against a fixed SWE-bench Lite smoke set, with doctor, run, and report commands.

NakshGuard is a self-hosted reverse proxy for OpenAI and Anthropic chat APIs that detects and blocks runaway AI agent loops before they waste tokens or API spend. It includes per-agent session tracking, configurable thresholds, shadow mode, and hot config reload.

A local-first, privacy-preserving usage meter for AI coding tools. It reads local transcripts to estimate token usage, cost, focus time, and achievements across tools like Claude Code, Codex, Kimi, and Qwen CLI.

Jo is a secure programming language designed to help catch prompt injection and other unsafe behavior at compile time by enforcing explicit capability boundaries. The repository shows the language’s security model, code examples, installation flow, and current project status.