The Open-Source AI Radar

Persistent memory layer across agent sessions with automatic semantic summaries and token-cost transparency.

CLI proxy that compresses shell command output before your AI coding assistant sees it, cutting tokens 60-90%.

AI coding skill that turns a folder of code, docs and images into a queryable knowledge graph across 20+ agents.

Lightweight personal AI agent for tools, chats and workflows - one binary, multi-channel.

Official Chrome DevTools MCP server giving coding agents control of a real Chrome instance with DevTools-grade inspection.

Python library from Google for LLM-powered structured extraction with source grounding.

Native desktop app for a GUI/computer-use agent powered by the open-weight UI-TARS model.

A free, on-device alternative to ElevenLabs for TTS, voice cloning and dictation.

Write HTML, render deterministic MP4 video - built for agents.

Tokenizer-free TTS from OpenBMB covering 30 languages with voice design and real-time streaming.

Framework of autonomous AI hacker agents for dynamic application security testing.

MIT-licensed open TTS with zero-shot voice cloning - 500M params, 23+ languages.

Agent-native personalized learning workspace - tutoring, quizzes, research and RAG knowledge bases.

Token-Oriented Object Notation - a compact serialization format that uses ~40% fewer tokens than JSON.

Crash-proof markdown planning for AI coding agents - persists task state across context loss and /clear.

Independent MCP server giving AI assistants structured access to n8n node docs to build and validate workflows.

In-page JavaScript GUI agent - control any webpage with natural language, no headless browser or extension.

Open-source Agent Operating System in Rust - autonomous capabilities, 40 channels, WASM sandbox.

macOS-native LLM inference server for Apple Silicon with continuous batching and SSD-tiered caching.

Self-healing browser harness that lets LLMs drive a real browser via CDP.

Memory framework for proactive AI agents - typed memory graph from chats, docs and media.

RAG on everything - graph-based vector index claiming 97% storage savings for private on-device search.

Code-search MCP server for AI coding agents - semantic + BM25 hybrid search across whole codebases.

Rust vector index with TurboQuant compression (ICLR 2026) - SIMD kernels, online ingest.

Secure, fast, extensible sandbox runtime for AI agents - multi-language SDKs, Docker/Kubernetes-native.

Fast on-device TTS via ONNX with 31-language support, running on CPU, browser and mobile.

High-accuracy document digitization (OCR/layout) with code and an open model.

Self-evolving memory OS for LLMs and AI agents with tiered L1-L3 memory.

Lightweight, in-process embedded vector database from Alibaba with multi-language bindings.

Provider-agnostic AI code-review CLI that emits line-level comments from Git diffs, open-sourced by Alibaba.

Self-hosted browser-control server that lets multiple agents drive Chrome via accessibility refs, not screenshots.

Native macOS video editor with an open editor core that AI agents drive directly through a local MCP server.

Text-audio foundation model - conversational TTS in 100+ languages with zero-shot cloning, 4B params.

Claude Code skills that turn an Obsidian vault into a self-organising AI second brain you fully own as Markdown.

Desktop agent IDE that runs 30+ coding agents side by side, each in its own isolated git worktree.

Lightweight Tauri/Rust desktop dev workspace bundling terminal, editor, Git and a local-or-BYOK AI agent.

Stealth headless browser for agents that wraps Camoufox for engine-level fingerprint spoofing, via a REST API.

Open-source meta-harness that orchestrates Claude Code, Codex, Cursor and custom agents behind one interface.

TypeScript framework that turns a goal into a runtime task DAG and parallelises multi-agent runs via one call.

Desktop app for local dictation, zero-shot voice cloning from a 3-second clip, and video dubbing, all on-device.

On-device TTS with instant voice cloning - ~360M active params, GGUF builds for phones, laptops and Raspberry Pi.

Expose the on-device Apple Intelligence model on macOS 26 as a zero-setup OpenAI-compatible local API.

AI-native, open-source Figma alternative with MCP and agent integration.

Opinionated, zero-config macOS terminal built on WezTerm with a built-in AI command-assist layer.

CPU-only semantic code search for agents: query in natural language, get back only the relevant snippets.

Pixel-native retrieval: renders documents as screenshots and searches over the images with a vision embedding model.

Pure-Rust local inference engine with an OpenAI-compatible API, shipped as one binary.

Open meta-harness that orchestrates Claude Code, Codex, Cursor and other coding agents behind one governed layer.

Agentic end-to-end video production system - 12 pipelines, brief to MP4.

Self-hostable platform for building production voice agents through a visual workflow builder.

Local, self-hosted long-term memory store modeled in sectors with a temporal knowledge graph.

Local desktop app that turns images into 3D models on your own GPU, no cloud, with GLB/OBJ/STL export.

0.1B multilingual TTS with zero-shot voice cloning that runs realtime on a CPU, fully open weights.

Local visual DevTools for Claude Code: inspect tool calls, token attribution, context and subagents from session logs.

Let AI agents control your existing Chrome via Playwright and MCP - logins and cookies intact.

Code-intelligence MCP server - a persistent knowledge graph over 158 languages via tree-sitter AST parsing.

Local-first AI notebook desktop app that pulls files and web content into notebooks with source-linked citations.

Open speech and sound generation family - nano to 8B, 31 languages, real-time streaming.

Unified virtual filesystem that mounts ~50 backends (S3, Drive, Slack, Gmail) as one filesystem for agents.

btop-style terminal dashboard for live monitoring of Claude Code, Codex and OpenCode agent sessions.

AI video-editing agent that turns natural-language directives into finished short videos, from the FireRed lab.

Local Rust 'context layer' that compresses file reads and shell output for coding agents over MCP.

CLI that lets AI agents drive and verify real iOS, Android, desktop and TV apps via accessibility-tree snapshots.

Human-led multi-agent orchestration where specialist agents persist as files in your repo and run in parallel.

MLX-native proxy that runs the Claude Code CLI against local models on Apple Silicon, fully offline.

Zero-setup, daemonless sandbox that runs coding agents under least-privilege via OS kernel allow-lists.

CLI that detects your hardware and ranks local LLMs that will run well on it, scored against real benchmarks.

Persistent personal agent server in Rust, one binary, with sandboxed tools, encrypted memory and many chat channels.

Self-hosted personal agent with markdown-defined personality, permission-gated tools and persistent memory, run from CLI/web/Telegram.

microVM sandbox that forks isolated Linux VMs for agents from a warm snapshot in milliseconds.

CLI that compiles documents into a cross-linked, Obsidian-friendly wiki instead of re-querying like classic RAG.

Local OpenAI-compatible inference server for Apple Silicon built on MLX, designed for coding agents.

AI-native state store for agents - MySQL-compatible hybrid vector + full-text search with FORK/MERGE sandboxes.

26M-parameter open-weights model for single-shot function calling on phones, watches and glasses.

CLI that fetches real dependency source code so coding agents can read internals, not just type stubs.

LLM CLI that turns unstructured text into typed knowledge: lists, tables, graphs, hypergraphs, via YAML templates.

Free push-to-talk voice input for macOS and Windows: hold a key, speak, get AI-polished text at your cursor.

Official CLI that lets coding agents run real end-to-end tests against a live app and get back structured failures.

Embedded AST-based semantic code-search CLI on the CocoIndex engine, with MCP and skill distribution.

One-command installer that turns your own machine into a private AI server (LLM, voice, RAG, image gen).

Turn any MCP, OpenAPI or GraphQL server into a CLI at runtime - zero codegen, token-efficient.

Cross-agent memory layer backed by human-readable Markdown and Milvus, local by default.

Python reliability layer that makes self-hosted and small LLMs dependable at tool-calling, with guardrails.

Open-source AI workbench that runs the full scientific research loop - literature, code, experiments, write-up - from a single goal.

Hosts an Apple Simulator over HTTP and ships an Agent Skill so coding agents can run and verify iOS apps.

Self-hostable chat-with-your-data agent: upload Excel/CSV or a database, ask in natural language, get SQL, charts and reports.

Open Vietnamese (plus English) TTS with instant voice cloning, trained from scratch and runnable on CPU.

On-device, real-time voice and vision AI - powered by Gemma and Kokoro, no cloud.

Self-hostable document-to-chunks layer for agentic RAG: parses PDFs and Office files into structured chunks with citations.

Credential broker that proxies agent traffic and swaps in real secrets, so the agent never sees them.

Rust object-storage engine on io_uring that moves data the shortest path from NVMe to GPU memory for training and inference.

Inference engine pairing a C++ scheduler with CUDA kernels for high-throughput agentic LLM serving.

MCP server that adds image and video generation (GPT Image 2, Seedance, ComfyUI) inside coding agents.

vLLM-style local server for Apple Silicon that speaks both the OpenAI and Anthropic APIs, with multimodal support.

Unsloth-compatible fine-tuning for LLMs, VLMs, TTS and STT on Apple Silicon via MLX.

Cross-agent memory layer with three verbs (remember, recall, answer) over an information-theoretic engine, no vector DB.

Background memory plugin for the OpenCode and Pi agents that replaces compaction with a self-managing historian.

MCP server that gives Claude security-intelligence tools, orchestrating NVD, EPSS, CISA KEV and more for CVE triage.

Japanese flow-matching TTS with zero-shot cloning and emoji-driven style control, MIT on code and weights.

On-device Apple Silicon speech toolkit (ASR, TTS, diarization, VAD) wiring 40+ open models via MLX.

Rust-core PDF toolkit with 7 language bindings - extraction, markdown conversion and an MCP server.

Rust tool that probes a CLI as a black box and scores how ready it is for AI agents to drive.

Fully offline durable memory for Claude Code - captures each session and generates a compact resume with no API calls.

Self-hosted WebUI workbench for GPT-image-2 with a local job queue and OpenAI-compatible auth.

Zero-setup desktop studio bundling local image generation, GGUF LLMs, TTS and STT.

A factory that scaffolds your own branded agent harness with CLI, MCP server and signed releases.

Agent skill that lets coding agents discover a running marimo notebook and execute code in the live session.

A native macOS terminal that organizes agent sessions into workspaces instead of tabs.

An MCP server that drives a real Linux desktop via accessibility APIs and input injection.

A Rust coding agent that drives Claude Code / Codex / OpenCode as a simulated multi-role dev team.

On-demand Firecracker microVMs you can hand to an AI agent, self-hosted on your own KVM box.

Local GUI for OpenAI Codex image workflows: pixel art, sprite sheets and frame animation via a file inbox/outbox handoff.

Self-hosted, bring-your-own-API image-generation workbench with an infinite canvas and PWA.

Local-first desktop app that turns documents into clean, AI-ready Markdown, OCR included.

Durable MCP control plane that turns Codex Desktop into a resumable async worker.

Persistent, shareable memory for AI coding agents across sessions and tools.

Free bidirectional Figma MCP server: designs to framework-aware code and back to the canvas.

Give any AI agent eyes and hands on one or more desktops via computer-use tools.