Open-source AI and automation projects

Maestro

Multi-agent system with 30+ specialized AI agents

Orchestration system coordinating 30+ purpose-built AI agents across research, coding, writing, analysis, and automation tasks. Agents collaborate in configurable pipelines with shared context and tool access, making complex multi-step workflows feel like a single conversation.

Quorum

Multi-agent orchestration platform with five execution patterns

Full-stack platform for building, running, and observing AI workflows composed of multiple cooperating agents. Supports sequential, parallel, hierarchical, consensus, and debate patterns via LangGraph, with real-time SSE streaming, human-in-the-loop approval gates, RAG memory, and a visual ReactFlow workflow UI.

OpenCode Agents System

79-agent system with 33 custom commands for AI-assisted development

Comprehensive agent framework for OpenCode with 79 specialized agents covering code review, testing, documentation, refactoring, security auditing, and architecture planning, plus 33 custom slash commands that compose agents into multi-step development workflows.

OpenWiki

AI-powered code documentation generator built in Rust

Parses codebases via tree-sitter (10+ languages), generates structured wiki pages with configurable LLM backends (Anthropic, OpenAI, Gemini, Ollama), and exposes indexed docs through a REST API, MCP server, CLI, and Next.js UI with hybrid full-text and vector search via Reciprocal Rank Fusion.

Vertex AI MCP Server

Google Cloud Vertex AI and Gemini integration, 20+ tools

MCP server exposing 20+ tools for Google Cloud Vertex AI and Gemini models, text generation, multimodal reasoning, image analysis, code generation, embeddings, and document processing. Supports streaming responses and all Gemini model variants.

Google AI Search MCP

Google AI-powered search tools for developers via MCP

MCP server providing AI-enhanced Google search capabilities for developer workflows, query expansion, result summarization, code snippet extraction, and documentation lookup. Current search results can ground an answer, but generated output still requires verification.

CodeVault

Semantic code search and MCP server for any codebase

Published npm package that indexes codebases using tree-sitter AST parsing and vector embeddings, exposing search, Q&A, and interactive chat through both a CLI and MCP server. Claude Desktop or any MCP client can query your entire codebase in natural language with hybrid BM25 + vector retrieval.

Google Ads MCP

AI-native Google Ads management via Model Context Protocol

MCP server exposing 17+ Google Ads API tools for AI agents, campaign creation, keyword research, ad copy generation, bid optimization, and conversion analysis. Ships 14 pre-built AI agent definitions for roles like keyword researcher, compliance auditor, and optimization strategist.

xAI MCP Server

Grok-powered MCP tools for search, code, and file workflows

30+ MCP tools wrapping xAI's Grok models for web-search-grounded Q&A, code analysis, security review, configurable regulatory checklists, and filesystem operations. Checklist output is guidance for human review, not a legal or compliance determination.

Apollo.io MCP Server

Sales intelligence and lead generation via MCP

MCP server wrapping the Apollo.io API with 20 tools across lead search, contact enrichment, account intelligence, sales sequences, and activity logging. Deploys to Cloudflare Workers or Node.js with documented n8n integration for AI workflow automation pipelines.

Search & RAGDocs MCP

40+ MCP tools for search, scraping, RAG, and Supabase

Unified MCP server aggregating Brave Search, Exa AI, Firecrawl, Tavily, Supabase management, and a RAG documentation pipeline in one interface. Supports OpenAI, Gemini, and Ollama embeddings plus asynchronous background processing when installed from the repository.

Gemini Load Balancer

API key rotation and load balancing for Gemini models

Deprecated Gemini proxy implementation with API-key rotation, quota tracking, failure detection, and an OpenAI-style chat-completions endpoint. Retained as an implementation reference; the newer multi-provider load balancer is the maintained successor.

Deprecated in favor of the maintained multi-provider LLM Load Balancer.

LLM Load Balancer

OpenAI-compatible load balancer across multiple AI providers

Provider-agnostic load balancer with an OpenAI-compatible API surface, routing requests across multiple backends with health checks, weighted round-robin, and automatic failover. Drop-in replacement for OpenAI SDK clients that need multi-provider resilience.

ModelGate

Self-hosted API gateway unifying multiple LLM providers

Go-based gateway that exposes a unified OpenAI/Claude/Gemini-compatible HTTP interface in front of multiple AI backends. Handles OAuth credential management, automatic token refresh, quota failover, and hot-reload config, with PostgreSQL, Git, and S3 persistence backends.

QuickScribe

Real-time AI dictation with OS-level text injection

macOS/Linux desktop app that captures microphone audio via keyboard trigger and streams LLM-processed, grammar-corrected text directly into any active application. Supports Groq, Gemini, OpenAI, and Anthropic with offline local transcription via Whisper and VOSK.