Making every AI work together locally.
Proxima began in late 2025 as a personal project. As a developer building complex software projects with AI coding agents, I ran into a major wall: API subscription fatigue and high key costs. I was paying for separate ChatGPT Plus, Claude Pro, and Perplexity subscriptions, yet running agent queries directly through raw developer API endpoints was costing hundreds of dollars a month on larger code bases.
On top of that, raw API endpoints were often returning outdated or hallucinated code segments due to fixed knowledge limits, while active web models had direct internet access and newer contextual knowledge.
Our mission is to democratize developer access to advanced reasoning models. By turning local interfaces and browser sessions into a secure, standard local Model Context Protocol (MCP) server, we give developers a privacy-first bridge to run multi-agent collaboration, consensus queries, and self-healing compilers without subscription traps or expensive cloud fees.
Proxima is designed around absolute user privacy. Every request, token intercept, and database vault is maintained entirely on your local machine (127.0.0.1). No centralized accounts are needed, no telemetry is uploaded to our servers, and your API keys are stored securely using your operating system's native hardware keychain vault (via Electron's SafeStorage API).
What started as a personal script has now grown into a comprehensive tool suite registered with Cursor, VS Code, Windsurf, and Claude Desktop, trusted by thousands of developers worldwide to run AI coding flows efficiently.
Proxima is built and maintained by Zen4-bit. If you find the tool helpful, please consider supporting the project by sponsoring us on GitHub or sharing feedback on our issues page.