maximhq/bifrost
Open-source, high-performance AI gateway that works as your LLM gateway, MCP gateway and agent gateway. One OpenAI-compatible API for 10,000+ models, with failover, load balancing, semantic caching, virtual keys, budgets and MCP Code Mode. Works with any SDK or agent and adds under 200µs of overhead at 5k RPS.
8 best points · 4 comments · 16 Hacker News posts · 9,416 GitHub stars · View on GitHub
Hacker News discussions
- Bitfrost – LLM gateway 90x faster than Litellm at p99 (8 points, 1 comments, by havercosine, 2025-08-07)
- 50× faster than LiteLLM: Bifrost is a Go-based LLM gateway built for scale. (5 points, 0 comments, by PranayBatta, 2025-12-08)
- Show HN: Bifrost – High-performance AI gateway (4 points, 0 comments, by aanthonymax, 2026-10-13)
- Show HN: Bifrost – open-source LLM Gateway (50x lower latency than LiteLLM) (3 points, 0 comments, by dskuldeep, 2025-12-09)
- Show HN: Changed how I use agent harnesses (3 points, 0 comments, by Swapnoneel, 2026-08-03)
- I Found One Interesting Open Source GitHub Gem (3 points, 0 comments, by aanthonymax, 2026-09-16)
- Every LLM gateway we tested failed at scale – ended up building Bifrost (2 points, 1 comments, by PranayBatta, 2025-12-10)
- Show HN: Fastest LLM gateway (50x faster than LiteLLM) (2 points, 0 comments, by aanthonymax, 2026-01-30)
- Show HN: Fastest Enterprise AI Gateway (2 points, 0 comments, by aanthonymax, 2026-07-02)
- Show HN: A model router should also provide a MCP gateway (2 points, 0 comments, by Swapnoneel, 2026-08-07)
- Show HN: Bifrost: Fastest enterprise AI gateway (1 points, 0 comments, by aanthonymax, 2026-02-13)
- Bifrost CLI and Codex CLI: One Command to Set Up OpenAI Agent with Any Model (1 points, 0 comments, by aanthonymax, 2026-03-19)
- Is Hermes Agent unusable without a model router? (1 points, 1 comments, by Swapnoneel, 2026-08-03)
- Show HN: Bifrost: Enterprise MCP Gateway with Built-In Security: OAuth 2.0, RBAC (1 points, 0 comments, by aanthonymax, 2026-08-05)
- I thought an LLM gateway was unnecessary. Then we added a second model provider (1 points, 1 comments, by Swapnoneel, 2026-08-10)
- The worst part of using multiple LLM providers is not the API differences (1 points, 0 comments, by Swapnoneel, 2026-08-11)