Code Puppy
An open-source, privacy-first AI coding agent supporting 65+ LLM providers, including fully local models, with zero telemetry.
🔗 Visit Code Puppy
Description
Most AI coding assistants lock you into one company's AI model and quietly send your code to their servers. Code Puppy is built for people who'd rather not do that — it's an open-source coding assistant that works with more than 65 different AI providers (including ones you can run entirely on your own machine) and sends zero usage data anywhere.
Code Puppy is an open-source, privacy-first AI coding agent supporting 65+ LLM providers (OpenAI, Anthropic, Google, Cerebras, Ollama, and more via models.dev), with custom agent creation, MCP server integration, and a round-robin model distribution feature to work around individual providers' rate limits. It has zero telemetry by design and gives users full control over which model runs their code — including fully local models via Ollama. It's MIT-licensed and actively maintained, positioned as a transparent alternative to commercial closed-source coding agents like Cursor or Windsurf that have raised prices or restricted model access.
💬 Our review
The short version: Code Puppy is what you'd build if you wanted Cursor's AI-coding-agent experience but refused to be locked into one vendor's models or send your usage data anywhere — it's open source, supports 65+ LLM providers, and has zero telemetry.
The provider flexibility is the real differentiator: you can point it at a fully local Ollama model for maximum privacy, or round-robin across multiple API keys to dodge rate limits, neither of which closed commercial tools typically allow.
The tradeoff is that you're managing your own API keys and configuration instead of a polished, one-click commercial product — Cursor and Windsurf still have larger communities, better IDE integration, and less setup friction. If vendor lock-in or data privacy around your code is a real concern, Code Puppy is a legitimate answer. If you just want the smoothest out-of-the-box experience, Cursor or Windsurf remain easier.
💰 Pricing
📊 Global score
🤖 AI-enriched data
Gratuit et open source (licence MIT). Vous fournissez vos propres clés API pour les fournisseurs LLM payants (ou utilisez un modèle local gratuit via Ollama).
Pros
Plus de 65 fournisseurs LLM supportés, y compris modèles locaux
Zéro télémétrie par conception
Répartition round-robin pour contourner les limites de taux
100% open source (licence MIT)
Cons
Gestion de ses propres clés API requise
Communauté plus petite que Cursor/Windsurf
Configuration initiale plus technique