Bifrost
Maxim AI's open-source Go gateway for routing, budgeting and caching LLM traffic
Overview
Bifrost is Maxim AI's open-source (Apache-2.0) LLM gateway, written in Go. It puts more than 20 model providers behind one OpenAI-compatible API, with automatic fallbacks, virtual-key budgets, semantic caching, OpenTelemetry metrics and an MCP gateway, self-hosted via npx, Docker or Kubernetes. A paid Enterprise licence adds guardrails, cluster mode, adaptive load balancing, SSO and in-VPC deployment.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 12 Oct 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- OpenAI-compatible drop-in replacement
- Budgets and rate limits with virtual keys
- Automatic fallbacks
- Simple and semantic caching
- MCP gateway with Code Mode
- Guardrails
- Cluster mode and adaptive load balancing
- SAML and OIDC SSO with RBAC
- Vault support, audit logs and log exports
- SLA-backed support
What you can produce with Bifrost
- One OpenAI-compatible API across 20+ providers and 1,000+ models
- Automatic provider and model fallbacks with custom routing rules
- Virtual keys with hierarchical budgets and rate limits
- Simple and semantic response caching
- OpenTelemetry, Prometheus metrics and a built-in request dashboard
- MCP gateway with Code Mode for tool-calling agents
- Enterprise guardrails, cluster mode, SSO and in-VPC deployment
Bifrost's open-source core is genuinely useful: fallbacks, budgets, caching, OpenTelemetry and an MCP gateway in a fast Go binary that starts in seconds. Its latency claims are Maxim's own and inconsistent across its pages, so run your own load test.
Several headline features, including guardrails, cluster mode and adaptive load balancing, are Enterprise-only and priced on request with a 14-day trial. A strong option for teams hitting the limits of a Python gateway.
What we found
Bifrost is the open-source LLM gateway from Maxim AI, the trading name of H3 Labs Inc. of Mountain View, California, co-founded by Vaibhavi Gangwar and Akshay Deo and backed by Elevation Capital. The gateway has quietly become the company's lead product: the getmaxim.ai homepage now presents Bifrost, while Maxim's original evaluation and observability platform has moved to its own section of the site.
The pitch is speed. Bifrost is written in Go, and Maxim's FAQ claims 11 microseconds of added overhead at 5,000 requests per second, with P99 latency more than fifty times lower than LiteLLM's in its own test. Those are vendor numbers, and they do not agree with each other: the homepage shows 20 microseconds and the GitHub description says under 100. Treat them as a reason to run your own load test, not as a result.
What is verifiable is the open-source core. The maximhq/bifrost repository is Apache-2.0, had fresh releases this month, and gives you one OpenAI-compatible endpoint for more than 20 providers, automatic fallbacks, virtual keys with budgets and rate limits, simple and semantic caching, OpenTelemetry and Prometheus metrics, and an MCP gateway with a Code Mode. It starts with a single npx or Docker command. Maxim says more than 1,000 teams use Bifrost, a figure it does not break down.
Some features often credited to Bifrost sit behind the Enterprise licence: guardrails, cluster mode, adaptive load balancing, SAML and OIDC single sign-on, role-based access, vault integrations, audit logs and in-VPC or air-gapped deployment. Enterprise is priced on request, with a 14-day free trial, and Maxim's trust centre lists SOC 2, ISO 27001, HIPAA and GDPR compliance at company level.
Independent scrutiny is still thin; its Hacker News submissions drew almost no discussion, whatever earlier versions of this page said. Teams outgrowing a Python gateway should shortlist it.
Compiled by RECATOOLS from public sources: vendor documentation and pricing, code repositories and published reporting.
About this listing
This entry was compiled from publicly available data including Bifrost's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Bifrost unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Bifrost directly →
Spotted something out of date? Suggest an update →
Related reading
Alternatives to Bifrost
More in LLMs & Chat
