open-source

Open-source AI sandbox infrastructure for running code, browsing the web, and isolating agent workflows inside disposable microVMs.

A self-hosted Docker Compose stack for running local AI services including Ollama, LiteLLM, Whisper, Kokoro, embeddings, and an MCP Gateway.

Open Bias is an open-source reliability harness for LLM apps and agents that enforces rules at runtime. It sits between your app and an LLM provider, using RULES.md policies to trace, block, or fix off-policy behavior.

Open-source voice AI from Microsoft with both long-form text-to-speech and speech recognition models. The repo highlights 90-minute multi-speaker TTS, 60-minute single-pass ASR, multilingual support, hotwording, and links to docs, Hugging Face, playground, finetuning, and papers.

prxhub is an open registry for AI research bundles, built around verifiable .prx artifacts, search, provenance, and agent-friendly publishing via MCP and CLI.

Release page for vLLM v0.28.0, a high-throughput and memory-efficient inference and serving engine for LLMs. The snapshot shows release highlights, model support updates, breaking changes, and installable artifacts for PyPI, Docker, ROCm, CPU, and XPU.

unbox-ai is an open-source CLI and local viewer for inspecting AI agent traces. It visualizes token usage, latency, tool calls, and repeated context in a bundle-analyzer-style interface, and also includes read-only commands for agents and a live AI SDK DevTools mode.

Open-source, self-hosted remote controller for coding agents like Claude Code, Codex, opencode, and Kimi Code. It runs sessions on your own Linux box or Mac with tmux underneath and lets you monitor and approve work from desktop, phone, or browser.

Open-source AI IDE for running coding agents in parallel with isolated git worktrees, reusable workflows, and self-hosting support.

Phinq is an open-source runtime governance layer for AI agents that intercepts tool calls, classifies risk, pauses irreversible actions for human approval, and records decisions in a tamper-evident audit log.

Open-source coding agent that keeps a persistent memory of your repo, with a CLI, self-hosted gateway, model routing, and support for OpenAI-compatible and provider-native endpoints.

A web-based tester for MCP servers that inspects tools, resources, and prompts, measures latency, estimates token usage, and flags tool-poisoning risks.

compactdiff snapshots, diffs, replays, and branches AI coding-agent sessions locally. It reads Claude Code and Codex logs natively, redacts sensitive data by default, and exports shareable replay HTML or forked JSONL branches.

Chrome extension that helps humans review AI-generated GitHub PRs by clustering diffs into review units and adding short AI summaries.

An AI-native 3D editor built on Blender with an embedded agent for 3D production tasks like retopology, UV rebuilds, baking, and export prep.

Open-source runtime guardrails for AI agents with cost limits, loop detection, policy enforcement, observability, and dashboard/CLI support.

Ami is a local, open-source AI agent that connects to apps and tools with your own tokens, keeps a universal to-do list, and runs busywork on your behalf. It includes copilot chat and graph memory, and runs on your machine with local data storage.

Open-source project showing how to run a 28.9M-parameter language model on an ESP32-S3 microcontroller, with on-device inference, firmware, wiring, training code, and measured results.

Mission control for AI coding agents, letting users run, watch, resume, and orchestrate tools like Claude Code, Codex, Copilot, and opencode from one terminal and live dashboard.

A provenance-aware memory layer for AI agents that stores user facts, interaction history, and third-party claims with source tracking and quarantine rules.

Headroom is an open-source context compression layer for AI agents. It compresses tool outputs, logs, files, RAG chunks, and conversation history before they reach the LLM, with support for a library, proxy, and MCP server.

An open-source Python tool that controls AI agent loops using real-time convergence signals instead of a fixed max-iterations cap, with rollback and adapters for popular agent frameworks.

Mindwalk is a local Go tool that replays Claude Code and Codex sessions on a 3D map of a codebase, showing where an agent searched, read, and edited files over time.

Open-source football management simulation where AI agents manage a club through MCP while humans watch the action in a terminal TUI. The repo includes a daemon, spectator console, calibration CLI, and detailed setup and command docs.