Trending Repos

Open-source personal finance app with an AI financial advisor: track your net worth, investments, ETFs, cash and debts on your own computer.

Real-time network traffic visualizer — see every connection your computer makes. Privacy-first, 100% local. / Visualiseur de trafic réseau en temps réel, 100% local.

Not another ChatGPT wrapper. This is a local-first tool that turns your notes, web pages, PDFs and audio/video into a personal knowledge graph — running entirely on your own machine.

clodfarm (say it out loud): a farm of Claude Code agents. Plant a mission, they split it into sub-agents, open the work, and pace themselves on each account's real 5-hour and weekly usage. Steer it from the Claude app.

Free, open-source anti-detect browser profile manager: per-profile fingerprint, cookies and proxy, Camoufox and Chrome engines, desktop app, REST API and MCP server for AI agents. Self-hosted alternative to GoLogin / MoreLogin.

Self-hosted audiobook discovery, Goodreads and Hardcover lists, Audiobookshelf integration, and automatic downloads.

Self-hosted live dashboard for the Xiaomi Mi Body Composition Scale 2 over Bluetooth Low Energy: profiles, body composition, history

A curated list of GEO and AI visibility tools, compared on which AI engines they track, how often they refresh, what they cost, and whether you can get your data out.

Local decision model with calibrated probabilities: send a state and yes/no, choice or score questions, get a probability for every option. 0.8B GGUF on CPU, Jev-style API.

ESPN draft and lineup tools for Codex and other MCP clients, with browser sign-in, approval modes, per-team limits, and server monitoring.

Benchmark local LLM serving by replaying real agent trajectories

LearnBuddy 学习伙伴：可私有部署的个性化 AI 学习系统，支持 Cloud 体验与 Personal BYOK，可选 OpenClaw + 飞书。

Self-hosted AI social media studio: one piece of material in, platform-native posts out. Format router, self-critique, pre-publish risk gate, real OAuth publishing. Web app + CLI + MCP server + agent skill.

Local-first voice memory assistant — capture speech, transcribe live, and ask a private RAG chatbot grounded in what you've actually said. No cloud required.

从手机、平板或电脑浏览器继续 Codex App 会话，兼容官方登录、API 与自定义模型服务。Self-hosted browser access to your existing Codex chats, with local/SSH sessions, approvals and notifications.

Amnezia VPN v5.0.3.0 — Amnezia VPN Client (Desktop+Mobile) Amnezia VPN is an open-source VPN client that enables users to easily deploy and manage their own self-hosted VPN server. Info page and download link. Third-party software; all rights belong to the original authors.

Jev-style parallel constrained decisions for any MLX model on Apple Silicon. Typed, schema-valid JSON in one forward pass.

Explore sleep, training, mood and nutrition together using local records, traceable calculations and optional MCP tools for your preferred AI client.

A nightly newspaper that writes itself: an agent runs the desks, the figures come from code, and it refuses to publish an edition with marks in it.

Nexus — execution infrastructure for AI agents. Host agents in the cloud or behind NAT; let users attach and authorize their own computers and Android phones. Deploy once, work across devices. Python SDK · MCP · Shared run context · Browser, terminal and files · Built-in chat UI · Model routing.

Local proxy that learns your app's typed LLM decisions and answers them with a Laya head. Jev and OpenAI compatible.

Teach a local AI what your camera sees and turn it into Home Assistant sensors — garage door open/closed, gate shut, car parked. 100% local, any CPU.

Open-source alternative to TypeSafe's Jev: a System One style model layer that gives typed, calibrated decisions from any open-weights LLM in one forward pass (HF + vLLM), with honest benchmarks

Self-hosted threat intelligence platform — feed aggregation, AI triage, MITRE ATT&CK coverage, and Sentinel-integrated detection engineering. Runs standalone or fully Azure-integrated.

1min.AI — Business SaaS listing. Visit link via SOFTGIT. Third-party; rights belong to original authors.

OpenJev: Open-source alternative to TypeSafe Jev. Typed probabilistic decision API (Choice, Noul, Score) powered by open LLMs & constrained logprob calibration.

An experimental local AI runtime. It relays one request through five small models with one job each, on a single 8 GB consumer GPU.

A 3D world where AI agents live, work and teach, and you can watch everything they do. Self-hosted, works with local models.

A sovereign, self-hosted AI companion with persistent memory, daily routines, photos, and voice notes. Enforces real boundaries in code, runs 24/7 on your own hardware (Linux, Android Termux, Windows), and keeps your data 100% private.

Self-hosted macro and markets terminal running entirely on free data sources. Built AI-first.

Free, open-source desktop SEO crawler and technical site audit engine for macOS, Windows, and Linux. 250+ technical SEO, GEO, and AI-search checks with no page cap.

Nythos Plus is a single-file, standard-library-only Python runtime built around one idea: reasoning effort is a resource, not a personality trait. Bardiya Shokeri

Run one local MLX model across Apple Silicon Macs on macOS and on Omarchy Linux, served as one OpenAI-compatible endpoint

A living AI civilisation you can watch: tiny creatures discover, build, teach and pass knowledge on, with any LLM in their minds.

Open-source full-stack AI agent with an embeddable React UI and FastAPI runtime. Self-host with streaming, pause/resume, durable state, and parallel multi-agent subtasks.

Rulith Inference, formerly Strix Llama. Serving a 125B MoE model fast on one AMD Strix Halo machine, on Windows: a patched llama.cpp, a manager, and a desktop app.

Run Jev-style typed decisions locally on your Mac with low RAM usage and fast responses

Qwen3.8-27B EXL3 (4.00 bpw) + DFlash2 speculative decoding for ExLlamaV3, validated at 262k context on a 24 GB RTX 3090

Serving a 125B MoE model fast on one AMD Strix Halo machine, on Windows: a patched llama.cpp, a manager, and a desktop app.

Evidence-first news workspace: compare official vs market narratives, click any AI claim to the exact source sentence, and track how your judgment changes. Self-hosted — one-command demo, no API keys.

An self-hosted AI platform that talks to your customers for you: it answers in messengers, email and web chat, and hands your team only the hard questions.

Bonsai 2 27B: 262k context (trained max) on 12 GB, 54 to 97 tok/s decode, 2x prefill, q8_0 KV recipe, grammar tool calls 9/9. Same ternary weights. Windows bundle + receipts.

Logic-based multi-LLM voice router for Home Assistant, combining native intents, local Qwen, GPU-aware desktop Ollama, and cloud AI with intelligent fallback.

Compute-for-memory research lab for running larger LLMs on memory-constrained consumer GPUs.

Small, calibrated decision models on your own machine: systemone-compatible local server, 6 agent skills, Claude Code guard, MCP tools. Weights on Hugging Face.

When Jev meets LLM -- Pair a small local LLM (System 2) with Jev model (System 1) to make small model more faster and accurate — OpenAI-compatible, private, and almost free to run in Kilo code and Cline.

Like Johnny, your local agent carries more than its head can hold — and keeps working for hours. Transparent proxy: verbatim archive, structured recall, segments, saved engine state.

Measure whether a back-office AI task can run on your own hardware: accuracy vs gold, p50/p95 latency, $/1k tasks, amortised hardware cost and break-even volume, plus an OpenAI-compatible router that sends each task to the cheapest model meeting your quality bar.

LocalBase — Your self-hosted backend, in one Python file. A lightweight Supabase alternative with SQLite databases, auto-generated REST APIs, a SQL editor, data grid, API keys, realtime WebSockets, backups, and a beautiful dashboard. No Docker, no Node, no cloud. Just python localbase.py.

Typed decisions (choice / score / yes-no) from local Qwen models on Apple Silicon. Probabilities come straight from the logits, no text generation. TypeSafe-compatible HTTP API, runs on MLX.

Run System One decision models from Hugging Face locally behind the Jev / System One API, Ollama style

AiiStream Q3.6: byte-identical MoE expert streaming on Apple Silicon — runs Qwen3.6-35B-A3B on a 16 GB Mac

BloomGauge (formerly Bloomkeeper): free Darkbloom manager and earnings dashboard for Apple Silicon Macs. Source-available core (PolyForm Shield). Keeps the best-paying model loaded, recovers stalls, no Terminal needed.

Offline-first RAG for drilling engineers: ingest well files, get answers with verbatim-verified citations, and generate a pre-spud offset-well risk brief with NPT cost and mitigations quoted from end-of-well reports. Runs with no network access and no model download.

Turn Safari web chats (DeepSeek, etc.) into a local OpenAI/Anthropic-compatible API — no API key, no cost. Drop-in for Claude Code / ccswitch.（把 Safari 网页对话（DeepSeek 等）包装成本地 OpenAI/Anthropic 兼容 API —— 不要 API key、不花钱，可直接接入 Claude Code / ccswitch）

Local-first book translation and LaTeX/PDF pipeline. Configurable MT languages; current ready-to-run routes EN→RU and prepared DE→RU OCR. MinerU, Hy-MT2, resumable jobs, portable agent skill. No mandatory cloud or polish. 26 optional expert skills included.

Qwen 3.8 Memory Hack: This 0.8B AI Steals a 51B Brain Locally! (Qwengram Setup) - Local Python engine and Ollama setup for Qwengram-0.8B with transferred N-gram memory.

Open-source, local Jev alternative: a System One decision layer with provably calibrated confidence (ECE 0.170→0.071). Typed decisions, runs offline, no API key, no waitlist.

JEV's alike open-source alternative Laya-MLX used as a "LLM-as-a-judge"

Offline-first, multi-language SAST with semantic data-flow analysis, deterministic security rules, SARIF/PDF reporting, Docker isolation, and optional local Ollama enrichment.

⚡ Foq — the FREE, local, open-source alternative to Jev. Typed System 1 decisions in ~25 ms — no waitlist, no cloud, no per-token cost. foq.fr

A directory of works discussed in the Launch80 discord server

Academic DOI canonical identifier resolver & BibTeX record generator (CrossRef)

Academic DOI canonical identifier resolver & BibTeX record generator (CrossRef)

Scientific paper methodology auditor identifying selection bias & p-hacking (Elicit)

Scientific paper methodology auditor identifying selection bias & p-hacking (Elicit)

Complex multi-faceted research prompt decomposer & synthesis DAG planner (Perplexity / Genspark)

Complex multi-faceted research prompt decomposer & synthesis DAG planner (Perplexity / Genspark)

Academic smart citation context auditor & credibility badge indexer (Scite.ai / Perplexity)

Academic smart citation context auditor & credibility badge indexer (Scite.ai / Perplexity)

Multi-study scientific assertion analyzer & consensus ratio calculator (Consensus.app / Elicit)

Multi-study scientific assertion analyzer & consensus ratio calculator (Consensus.app / Elicit)

Automated promo code & coupon cascade tester for checkout discounts (Honey/Perplexity)

Verified merchant review citation synthesizer & grounded pros/cons (Perplexity)

Agentic one-click instant buy token broker & headless settlement (Perplexity Buy)

Conversational multi-turn gift & aesthetic style consultation matcher (Claude)

Claude Computer Use storefront cart action sequence & HITL approval planner

Qwen3.8-Flash-Next 177B on RTX 5090 Laptop (24GB VRAM + 64GB RAM): quant selection, three-tier memory split, 256K context, KV quantization (+3.4-5%) at 22-25 tok/s / 本地部署实录（中英双语）

Run Qwen3.8-Flash-Next on ONE RTX 3090 (24 GB) + 64 GB RAM: 128K context, up to 2,100 tok/s prefill, 43–51 tok/s decode. vLLM runtime with hot MoE experts on the GPU and cold experts computed on the CPU, INT8 KV cache, Docker, OpenAI-compatible API.

Typed decisions from local LLMs in one forward pass, via Ollama. No generation loop, no JSON parsing.

A Power BI specialist LLM for Ollama — local-first, grounded, 12/12 on a 12-task Power BI benchmark at ~156 tok/s. Built on gemma4:26b with a baked-in Power BI rulebook (DAX, M, PBIR, TMDL, safety gates) plus a unified RAG + gated-tools wrapper. PolyForm Noncommercial.

Router and dashboard for haloge-flash-server

Qwen3.8-27B on a single 24GB laptop GPU - from 53 community quants to the true hardware limit: self-built CUDA 13.3 engine, 256K (262,144-token) context with q4_0 KV + vision at 82 tok/s, one-line launch.

Local browser AI agent with indexed action space — Laya local decisions, /opt/homebrew/bin/bash/run. Google Flights demo in ~6.5s. Python.

Genetic algorithm optimizer with tournament selection, single-point crossover, bit-flip mutation, and elitism.

Genetic algorithm optimizer with tournament selection, single-point crossover, bit-flip mutation, and elitism.