Explore the directory

This project documents, step by step, how to assemble a self-hosted local LLM stack from open-source components: a chat interface, document embedding, a vector database, web search and several MCP servers.

• Tutorials

This is a Visual Studio extension that adds a chat panel for talking to local or OpenAI-compatible language models directly inside the IDE, with access to solution files and directories and the ability to generate new files.

• Coding agents

Lantor is a local-first, macOS desktop workspace where a user keeps multiple coding agents such as Codex and Claude warm, dispatches work to them through chat, and keeps every thread, task and attachment on their own machine.

• Coding agents

JW MCP Server is an MCP server exposing JW.org content - Bible scripture lookup with study notes, workbook materials, Watchtower articles and video captions - to AI assistants such as Claude, ChatGPT and Grok.

• MCP servers

iamnotacoder is a toolkit of LLM-assisted scripts that discovers Python repositories on GitHub, then improves their code style, fixes issues and adds tests, integrating with GitHub to fork and commit the results.

• Code quality

Dug applies semantic web and knowledge graph techniques to biomedical research metadata, letting a researcher search underused datasets such as dbGaP by concept rather than exact keyword, based on curated ontologies.

• RAG
• Data engineering

dsh-mcp-connector is a management panel for the DeepSeek Harness agent tool that lets a user connect, search and troubleshoot Model Context Protocol servers from a catalog, with OAuth 2.0, API key and JSON import support.

• MCP servers

DSA Context Engine gives AI coding agents a persistent memory that unifies code structure, tasks and past conversations into a single graph, served locally to Claude Code, Codex and other agent CLIs over MCP.

• MCP servers
• Agent memory

CodexA is a local semantic code intelligence CLI that indexes a codebase, builds embeddings and a BM25 index, and exposes structured search, retrieval-augmented answers and repository analysis to AI coding assistants over HTTP, CLI or MCP.

• MCP servers
• RAG

This kit is a pre-configured starter repository of agents, skills, slash commands and MCP integrations meant to bring a team onto Claude Code within about thirty minutes for a specific technology stack.

• Agent skills

cc-rig generates a Claude Code project setup - CLAUDE.md, hooks, agents, skills, commands and plugins tuned to a chosen framework - and then reports over time how much cache discipline it is actually saving.

• Agent skills

BlueRock is a lightweight runtime security sensor for Python processes and MCP servers, recording tool calls, session lifecycle events, module imports and subprocess execution as structured logs with no code changes.

• MCP servers
• Security

awesome-free-ai-coding is a self-updating, machine-verified list of coding agents and LLM API providers that offer a free tier, tracking which models are available for free and where.

• Coding agents
• Curated lists

AMA is a research system for AI agent memory that uses four collaborating agents and three memory granularities to store, verify and refresh what a conversational agent remembers over time.

• Agent memory

AILANG is a purely functional, effect-typed programming language designed as a deterministic execution substrate for AI-generated code, with reflectable and verifiable semantics for every construct.

• Languages

A Nix library for embedding, querying and stripping metadata on Nix expressions, storing arbitrary data such as a package's description, license or maintainers under a _meta attribute and locating the right binding to attach it to.

• Libraries

A work-in-progress, multi-paradigm compiled programming language designed to be simple for a large language model to write and easy for a person to read, aiming for performance within about 15% of C.

• Languages

A demo project that packages a self-hosted LLM chat interface, running Qwen2-1.5B-Instruct or another Hugging Face model behind vLLM with a Streamlit front end, into a single Docker container deployable locally or on an AWS GPU instance.

• Inference
• Tutorials

A multi-tenant vector database and key-value store in one process that speaks the Redis protocol, keeping full-precision vectors on SSD with only a small quantized working set in RAM.

• Databases

A Tour of Agents is an interactive, in-browser course that builds an AI agent from scratch in about sixty lines of plain Python across nine lessons, showing what agent frameworks do under the hood.

• Tutorials

Tightwad pools CUDA and ROCm GPUs across separate machines and speeds up local LLM inference with application-layer speculative decoding, sitting as a drop-in proxy in front of an existing model endpoint.

• Inference

t0-1 is a research demonstration from the Alan Turing Institute's t0 initiative, applying retrieval-augmented reasoning with lean language models to a domain-specific knowledge base built from NHS condition pages.

• RAG

Shebe is a fast, local BM25 full-text search tool for cloned code repositories, exposed to AI coding agents through an MCP server and a CLI, with no embeddings, GPU or cloud dependency.

• MCP servers
• CLI tools

RemembrallMCP gives AI coding agents whole-codebase knowledge: a field-aware dependency graph of functions, classes, methods and fields, plus persistent memory that survives between sessions, exposed to agents over MCP.

• MCP servers
• Agent memory