Bifrost

What it does

Bifrost is an AI gateway that takes application requests in an OpenAI-compatible format and routes them to more than 20 model providers, including OpenAI, Anthropic, AWS Bedrock, and Google Vertex. It returns model responses along with routing, caching, and governance applied to that traffic. Bifrost Edge extends this control to employee laptops by routing AI traffic from desktop apps, browser AI, and coding tools through the gateway.

Developers typically point an existing SDK at the gateway's base URL, then use virtual keys, budgets, rate limits, fallbacks, and load balancing per consumer. Platform teams can add observability through Prometheus, OpenTelemetry, and a Datadog connector, and can control which MCP tools are exposed to each key. Edge is a separate enterprise service that runs on macOS, Windows, and Linux and signs in through an existing identity provider.

The open-source gateway suits developers and small teams who can self-manage deployments via Docker, NPX, or the Go binary. Edge and the enterprise tier are aimed at infrastructure, security, and platform teams at larger organizations. Edge is currently in early access for existing enterprise customers, and full pricing has not been published. Enterprise features such as guardrails, clustering, and vault support require a license request.

What it can do

• Route requests from one OpenAI-compatible API to 20+ providers with automatic failover and load balancing
• Govern access per consumer with virtual keys, budgets, and rate limits
• Cache semantically similar responses to reduce cost and latency
• Expose and filter MCP tools, including an MCP gateway and Code Mode
• Export Prometheus metrics and OpenTelemetry traces for monitoring
• Route AI traffic from employee laptops through the gateway with Bifrost Edge (enterprise, early access)

Works with

• OpenAI SDK
• Anthropic SDK
• AWS Bedrock SDK
• Google GenAI SDK
• LiteLLM
• LangChain
• PydanticAI
• Datadog
• Okta
• Microsoft Entra

Before you choose

• Edge is currently in early access for existing enterprise customers; treat availability and pricing as provisional.
• The unversioned Docker commands are for local evaluation; production deployments need a pinned version, TLS, authentication, and durable storage.
• Enterprise features (guardrails, clustering, SAML/OIDC SSO, vault, audit logs) are not included in the open-source tier.

Listed under

• Coding
• AI governance
• AI routing
• API gateway
• Enterprise AI
• LLM gateway
• Open source

Get started

Edge requires registering for early access. Gateway can be run locally with NPX or Docker.

1. Start the gateway locally with npx -y @maximhq/bifrost or docker run -p 8080:8080 maximhq/bifrost, then open http://localhost:8080.
2. Before exposing Bifrost beyond your machine, set BIFROST_SETUP_TOKEN and enable dashboard password protection under Workspace → Config → Security.
3. For Edge, register for early access, then deploy via device management tools such as Jamf, Intune, or Kandji.

Edge is priced by device count, but no rate is published; existing enterprise customers currently receive it at no cost during early access. Enterprise requires a license request or a conversation with the vendor, and the pricing page mentions a 14-day free trial of Bifrost Enterprise. Model provider usage costs are billed separately by each provider.

Source: getmaxim.ai/pricing, checked 29 Oct 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.

• Bifrost AI Gateway - Bifrost ↗docs.getbifrost.ai
• Bifrost Edge + Gateway | Route, Govern, and Secure AI Traffic ↗www.getmaxim.ai
• Setting Up - Bifrost ↗docs.getbifrost.ai
• Pricing | Bifrost OSS and Enterprise ↗www.getmaxim.ai

Company website

getmaxim.ai: product, pricing and policy pages

YC directory mirror

Batch and company name

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

manual_review

Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.