Explore the directory

An LLM post-training framework for reinforcement-learning scaling, built on slime's training and data-generation design, that uses vLLM with a router as its default rollout backend.

• Models

A self-hosted, single-conversation AI agent framework that keeps one continuing conversation thread instead of separate chats, using asynchronous processing and self-directed context management to build long-term continuity and memory.

• Agent frameworks
• Agent memory

A one-click theme and skin switcher for the OpenAI Codex Desktop app on macOS and Windows, applying a custom look through Chrome DevTools Protocol injection without modifying the application itself.

• Coding agents
• Desktop utilities

A collection of 54 portable skills for Claude Code, Codex, OpenCode and Pi, covering code review, security audit, feature development, frontend design, project memory and authoring tools, plus an opt-in native sandbox for safer command execution.

• Agent skills
• Security

A single Rust binary that plans, applies and rolls back changes to a production ontology, reports every consequence of a change before it is applied, and produces a Lean 4 proof of the result that a reviewer can re-check without trusting the tool.

• Databases

A single Rust binary that builds a live knowledge graph and MCP server for a codebase, so Claude, Cursor and other AI agents can search it semantically and navigate its imports, calls and dependencies as structure rather than flat text.

• MCP servers
• Agent memory

MLX42 is a cross-platform, minimal windowing and graphics library that lets you draw and update textures on a window without touching the operating system's native windowing APIs directly.

• Libraries

An installable catalog of .NET skills and a CLI that installs them for Codex, Claude Code, GitHub Copilot and Gemini, so an AI agent uses current ASP.NET Core, Blazor, Orleans, MAUI or Semantic Kernel patterns instead of outdated ones.

• Agent skills

A curated list of developer tools, SDKs, libraries, testing utilities, proxies, hosting options and templates for building Model Context Protocol servers, organized by programming language and purpose.

• MCP servers
• Curated lists

A structural memory layer for AI coding agents that turns a codebase into a live, bi-temporal knowledge graph of functions, classes and call edges, queryable in milliseconds without an LLM call.

• Agent memory

A single binary under 100 MB that turns a Linux host into a self-hosted sandbox platform for running secure, isolated AI-agent workloads at scale, with gVisor, Kata or Firecracker isolation and an E2B-compatible API.

• Containers
• Hosting

A config-driven knowledge-distillation toolkit that turns a black-box teacher model into training data, composing generation, evaluation, filtering, rewriting, balancing and preference-scoring operators into pipelines that build SFT and DPO datasets.

• Models

An open-source course that builds a Claude Code-style coding agent from scratch in Python, covering the tool-calling loop, permissions, sandboxing, memory, session recording and a subagent fan-out.

• Coding agents
• Tutorials

A lightweight, cross-environment general-action agent that completes real digital tasks spanning a browser, desktop applications and the command line, coordinated by a TeamLeader that dispatches to specialized Workers.

• Agent frameworks
• Browser automation

A local-first control plane that runs Codex CLI, Claude Code, Cursor CLI, OpenCode and other coding-agent harnesses behind one typed interface, adding best-of-N races between harnesses, quota-aware account rotation and audited cost and quota tracking.

• Coding agents

A dual-architecture research framework that unifies the exact output of autoregressive large language models with the parallel, high-speed token generation of diffusion models, for strictly lossless faster inference.

• Models
• Inference

The official MCP server for Kagi Search, exposing web, news, video, podcast and image search with page extraction, filters and Kagi lenses, plus full-page-to-markdown extraction, to MCP-compatible clients.

• MCP servers

Upjet is a code-generation framework and Kubernetes controller runtime that turns any Terraform provider into a Crossplane provider, generating CRDs, reconciliation controllers and example manifests.

• Hosting

The official Model Context Protocol server for Salesforce DX, letting large language models read, manage and operate Salesforce org resources securely through a configurable set of tools and toolsets.

• MCP servers

A general-purpose AI agent framework and app for building, running and composing self-contained agents and multi-agent teams, with built-in tools, sub-agents, persistent sessions, a TUI and a web UI.

• Agent frameworks

A local-first service that gives AI agents "playbooks": portable sets of MCP tools, prompts and configuration that can be switched in and out of context and connected to Claude, Cursor or VS Code with one click.

• Agent frameworks
• MCP servers

A tool that exposes a Supermemory memory container as a real local filesystem or a virtual bash tool, so an agent or a person can read, write and grep memory the way they would any local directory.

• Agent memory

A modular reference application in which several AI agents, orchestrated by LangChain.js, LlamaIndex.TS and Microsoft Agent Framework, handle travel-agency customer queries, destination recommendations and itinerary planning through MCP servers.

• Agent frameworks
• MCP servers

A self-hosted service that converts a web URL, or as of v3 a PDF, Office document, EPUB, image, audio file or YouTube video, into clean Markdown for humans and AI agents to read.

• RAG
• Browser automation