Vani.ai: Unified AI workspace & API gateway for 34+ models (Claude, GPT, Gemini, Grok) in one subscription with 1M context, cited research, image/video gen, workflows, persistent projects & skills.
We built Vani AI out of frustration with juggling dozens of fragmented AI subscriptions, disjointed SDKs, and inflated token pricing. As developers and builders, we wanted one unified hub: instant access to 34+ frontier models (Claude, GPT, Gemini, Grok), seamless context switching, and real workflow automation via MCP. So we engineered both: an all-in-one conversational workspace and a drop-in API gateway that cuts inference costs by 30% without changing your existing OpenAI or Anthropic codebases. Built by developers, tuned for builders.
Our stack
• Core & Gateway: Rust and Go for high-throughput, low-latency API proxying; Python for agentic orchestration.
• Frontend: Next.js, React, TypeScript, and Tailwind CSS for a snappy, responsive workspace UI.
• Protocols & Standards: Model Context Protocol (MCP), SSE streaming, and OpenAI/Anthropic drop-in wire formats.
• Infrastructure: Kubernetes, Docker, Redis for high-speed rate-limiting, and dedicated GPU clusters with prompt caching.
• Data & Memory: PostgreSQL with pgvector for persistent project memory, context retrieval, and vector embeddings.