Top Picks

An open-source Claude Code skill set for academic research, covering research, writing, review, revision, and finalization. It includes guided workflows, installation steps, and pipeline documentation for researchers using AI in a human-in-the-loop process.

UltraCompress is a Python-based compression tool for large language models. The repo describes lossless 5-bit transformer compression, verification via SHA-256, a CLI on PyPI, and published model packs on Hugging Face.

Local-first long-term memory engine for AI agents with MCP, HTTP API, CLI, and Python usage. The page shows installation paths, IDE integrations, ingestion sources, and a live feature set focused on durable context, retrieval, and privacy.

A Docker-based sandbox for running AI coding agents with dropped capabilities, no-new-privileges, and no Docker socket access. It wraps the pi terminal coding agent and supports local file edits, persistent auth, and skills mounts.

CacheCore is an agent infrastructure tool that sits between LLM clients and the API to cache exact and near-identical prompts, cutting repeat calls and latency. The page shows OpenAI-compatible integration, support for common agent frameworks, and use cases for RAG and multi-agent workflows.

AI assistant for chatting with PDFs, helping users edit, fill, and understand forms step by step. The demo page shows a loaded IRS W-9 and supports guided form-filling workflows.

An autonomous neuro-symbolic formal verification engine for Lean 4. The repository describes a proof workflow that generates tactics and then checks them deterministically in Lean, with example proofs, verification steps, and telemetry outputs.

AI-powered trade signal product for retail traders that turns financial news into ranked trade plans with entries, stops, targets, and tracked accuracy.

An open-source Llama2 inference engine written in x86 real-mode assembly that fits in 1277 bytes and can boot directly from disk before any OS loads.

A Microsoft Word task pane add-in that embeds a Pi Agent-powered assistant for M365 Word, with production and local sideloading manifests plus build and test instructions.

Gate is an agentic desktop OS for developers that runs AI desks locally, extracts reusable skills from completed work, and logs every action with receipts. The page shows install flow, platform limits, model requirements, sandboxing, and a clear comparison with Cursor/Aider.

Tendril is an agentic sandbox that lets a model discover, build, register, and reuse tools across sessions. The repository describes its Strands-based agent loop, capability registry, Deno sandbox execution, and Tauri desktop shell.

An AI copilot for Autodesk Fusion 360 that installs as a Fusion add-in and lets users chat with the Adam palette inside the CAD app.

FastAPI backend for the Brush Up tutor app. It retrieves from a markdown note corpus, expands context through graph neighbors, and calls Anthropic to generate grounded programming help for Python, Ruby, and JavaScript.

GitGalaxy is a Python-based static analysis and knowledge graph engine for deep repository intelligence. It maps codebases at the function level, supports 50+ languages, runs locally, and exports JSON and SQLite for querying and AI-agent workflows.

Chrome extension that uses Gemini Nano via Chrome’s Prompt API to rewrite comments and posts live in a chosen tone, fully on-device with no API keys or backend.

A Claude Code playbook for engineers who want to drive internal adoption on their team, with guidance on what to share, how to answer questions, and a 30-day rollout plan.

AgInTiFlow is a web-first coding agent and CLI for controlled website automation, persistent conversations, resumable runs, and guarded local commands. The npm page shows quick-start install steps, model routing options, sandbox and browser tooling, session handling, and web UI support.

A workbench report comparing MiniMax M3 and GLM 5.2 on autonomous coding tasks, with scored results, latency and cost data, task-type breakdowns, and examples of where each model performed better.

GRID is a small framework for working with coding agents in a spec-driven way. The repository explains its purpose, basic workflow, and how it can be used in CI/CD to check for spec-code-test mismatches.

A VS Code blog post explaining how GitHub Copilot’s agentic harness is being optimized to reduce token usage, latency, and cost through extended prompt caching, tool search, and WebSocket connections.

A sourced model and short report exploring whether Europe could train a sovereign frontier AI model using public compute it already owns, with reproducible code, datasets, and a PDF report.

Foyer is a local dashboard for watching AI coding agents work, with a narrated current-focus view and a research panel for reading sourced briefings while you wait.

A macOS SwiftBar plugin that shows Claude Code quota gauges in the menu bar, with per-account usage, reset timers, and a dropdown with detailed limits.