Top Picks

A Node.js proxy server that rotates API keys for LLM APIs like OpenAI and Gemini when rate limits are hit. It also includes an admin panel, request logging, provider management, and a Telegram bot for chatting with configured models.

Lightweight AI inference engine written in Rust for embedding local LLMs in mobile and desktop apps, with bindings for Python, Flutter, and React Native.

A hands-on workshop repository that walks developers through building a small GPT-style language model from scratch, including tokenization, transformer blocks, training, and text generation.

A research report on the current MCP ecosystem, with live crawl numbers, verification rates, category breakdowns, and examples of both strong and weak MCP-positive sites.

TreasuryFlow is an AI CFO layer for businesses, aimed especially at teams moving off Bench. It connects to bank accounts via Plaid to show live cash balances, generate a 13-week forecast, and flag spending anomalies, while running alongside a separate bookkeeper.

AgentsDock is an IDE for agentic AI research that lets users connect Claude Code, Codex, and Cursor to their own machine and access them from desktop or mobile. The page shows setup steps, supported tools, and a clear workflow for remote agent research and model training.

Open-source LLM uncertainty and hallucination detection tool that estimates epistemic risk in microseconds using exact-match normalized entropy, with Rust gateway and Python package support.

A queue-driven battle league for AI agents with an MCP server, OAuth/sign-in support, and both live brawl and rap-battle modes. The page documents client setup paths, transport details, and how agents join matches and fight.

BiNeuron is an open-source local AI assistant for code generation, file analysis, and document/OCR handling. It detects programming languages, selects models based on hardware, and can run offline with proxy support when needed.

Ridge is a resource mesh for AI agents, with MCP, CLI, and Python access to local projects, Docker, SSH machines, and S3 storage. It focuses on delegated resource access, job coordination, and keeping artifacts out of model context.

Waiting Club is a group chat app for people waiting on AI responses, with desktop downloads for Mac and Windows, social features, mini games, and waiting-time tracking.

Botbin is a pastebin-style publishing tool for AI agent artifacts. It lets agents upload single-file HTML outputs like dashboards, charts, and visual diffs and share them through clean URLs instead of raw HTML in chat.

A research repo that replays real Claude Code and Mooncake traces to test prefix-cache eviction policies and compare them against LRU.

A capability-based microkernel written in SPARK Ada and formally verified with gnatprove, with a thin C SDK for building services and distribution components.

Bestie is an open-source native terminal coding agent harness focused on visibility and control, with sandboxing, provider support, session save/load, and cross-platform installs.

A detailed vLLM blog post explaining speculative decoding on AMD GPUs, with mechanics, supported drafting methods, configuration guidance, tuning advice, and benchmark discussion.

vernLLM is a TypeScript LLM call framework that keeps retries, fallback, rate limiting, circuit breaking, caching, and middleware inside your own process. It supports OpenAI-compatible providers plus Anthropic, Gemini, Bedrock, and others via adapters.

An open-source desktop app that watches your screenshot folders and uses an AI model to rename screenshots with meaningful filenames.

Rubato is a retro-Mac desk companion that mirrors AI coding session state and turns long waits into gentle break reminders. The repo includes firmware, plugin support, hardware details, and a clear explanation of how the device works.

A rented Linux box for AI agents to use real apps like Slack, Telegram, Discord, and the web through a private encrypted tunnel, with consent gating for outbound actions.

Loopback MITM proxy for Claude Code on Amazon Bedrock that can sniff, hold, edit, forward, or drop requests. The repo includes prerequisites, a quick start, flags, cost notes, and behavior limits.

React hooks for exposing app functionality as WebMCP tools, with Zod-based inputs, SSR-safe behavior, StrictMode safety, and a bridge to AI clients like Claude and Cursor.

An arXiv paper describing an open library of 163 procedural skills for research agents across 16 scientific areas, with versioned instruction files and accompanying reference material or runnable scripts.

Anthropic’s release notes for Claude Fable 5.1, covering new features, breaking changes, pricing, model availability, and migration details for API users.