LocalGPT
Open-source, local-first AI assistant with persistent markdown memory, multi-LLM-provider support, and natural-language 3D world generation.
🔗 Visit LocalGPT
Description
Most AI assistants live in the cloud and require an internet connection and a subscription, which means every conversation and every file you share passes through someone else's servers. LocalGPT is an open-source project that flips that: it's a local-first AI assistant that runs on your own machine, keeps a persistent memory stored as plain markdown files you can read and edit yourself, and works with whichever LLM provider you choose — Ollama or LM Studio for fully local models, or Anthropic and OpenAI if you want cloud-grade quality while keeping the assistant layer itself local.
The standout feature is LocalGPT Gen, which builds 3D worlds from natural-language descriptions — generating parametric shapes, PBR materials, lighting, and behaviors, then exporting the result as glTF/GLB files usable in other 3D tools. The core assistant supports autonomous task execution, hybrid web search, and SQLite-backed semantic search with local embeddings, all installable as a single binary with no Node.js, Docker, or Python dependency chain to wrestle with. It's available as a CLI, web UI, desktop app, or Telegram bot, runs sandboxed on Linux/macOS for security, and is licensed Apache-2.0 — free to use, modify, and self-host.
💬 Our review
The short version: LocalGPT is a genuinely interesting open-source project for developers who want a local, privacy-respecting AI assistant with an unusual bonus — natural-language 3D world generation — but it's a small project (1.1k GitHub stars, limited issue activity) next to better-funded local-AI alternatives.
Against mainstream cloud assistants like ChatGPT or Claude's own apps, LocalGPT's whole pitch is local-first privacy and multi-provider flexibility, at the cost of the polish and reliability that comes from a well-resourced commercial product. Against other local-AI projects like Open WebUI or LM Studio's own assistant features, LocalGPT differentiates mainly through its persistent markdown memory (which is refreshingly inspectable — you can literally read what the assistant 'remembers') and the 3D generation module, which is genuinely unusual and not something competitors offer. The single-binary install is a real usability win over projects that require a full Docker/Python stack. Worth trying if you want a local assistant and are curious about the 3D tooling; the smaller community means less polish and fewer integrations than you'd get from a bigger project.
💰 Pricing
📊 Global score
🤖 AI-enriched data
Open-source (licence Apache-2.0), gratuit, auto-hébergé. Nécessite votre propre fournisseur LLM (Ollama/LM Studio gratuits en local, ou clé API Anthropic/OpenAI payante si utilisé).
Pros
Fonctionnement 100% local, mémoire persistante lisible en markdown
Compatible multi-fournisseurs LLM (Ollama, LM Studio, Anthropic, OpenAI)
Installation binaire unique, sans Docker ni Python
Génération de mondes 3D par langage naturel, export glTF/GLB
Cons
Adoption encore limitée (1.1k étoiles GitHub)
Activité de maintenance modeste (peu d'issues traitées)
Moins de polish qu'un produit commercial financé