Skip to content
AgentsIndex

Bifrost by Maxim AI

Go gateway that routes model calls across many providers and governs agent tool access through the same endpoint.

Best for
Maxim AI is best for AI teams who need to test, monitor, and ship agents with tighter quality control.

What we checked

Pages last read
Facts checked
4 facts

From 4 sources, read 2 November 2026

AI answers
Too few read

58 buyer-question answers read, November 2026

At a glance

Tags
Freemium, Free trial, HIPAA

For developers

Works with
Langchain, LangGraph, OpenAI, OpenAI Agents, LiveKit, Crew AI and 8 more

Bifrost by Maxim AI in our research

All research

Study

AI Tool Pricing Index, November 2026

Pricing

Contact sales

82 of the 361 tools we checked ask you to contact sales for a price.

Edition price, checked 2 November 2026.

Pricing pages of the AI tools, November 2026 editionas of n = 361 tools checked

What is Bifrost by Maxim AI?

Maxim AI is an AI workflow platform for teams that test prompts and agents, simulate behavior, and monitor production runs. It combines Prompt Playground, prompt versioning, agent simulation and evaluation, and agent observability to tighten quality before and after release. The platform also includes Bifrost, a gateway for 1000+ models across 8+ providers, and is used by EY, ByteDance, Babylist, Klaviyo, and McAfee. Its published pricing covers Bifrost: a Free Forever OSS plan for self-managed deployments and a custom-priced Enterprise plan.

getmaxim.ai now sells Bifrost, an open-source AI gateway

The homepage opens on Bifrost, "The Fastest Enterprise AI Gateway", started locally with npx @maximhq/bifrost. It sits between an application and its model providers: an OpenAI, Anthropic, LangChain or Vercel AI SDK client changes its base URL and its requests then pass through Bifrost, which adds provider fallback, virtual keys with their own budgets and rate limits, routing rules, simple and semantic caching, and built-in observability with OpenTelemetry and Prometheus output. The docs overview offers two ways to run it: an HTTP gateway "with a built-in web UI for visual configuration and real-time monitoring", or a Go SDK embedded in a Go application.

The pricing page also lists a Bifrost CLI to "Govern coding agents like Claude Code, Codex CLI, and Gemini CLI through Bifrost", a prompt repository with versioned prompts, a mocker plugin for testing and asynchronous inference. The company's older product still shows through: the About page says "we are building an enterprise-grade AI evaluation and observability platform", and the pricing comparison carries a "Maxim AI Integration" row for "Optional Maxim SDK integration for LLM observability, tracing, and evaluation". The evaluation platform has no page of its own in the site navigation, and this listing covers Bifrost.

The MCP gateway gives agents one governed endpoint for tools

The MCP Gateway page says Bifrost "acts as both an MCP client (connecting to external tool servers) and an MCP server (exposing tools to external clients like Claude Desktop) through a single deployment." It connects to MCP servers over STDIO, HTTP or SSE, discovers their tools, and republishes them at one endpoint, where "Claude Desktop, Claude Code and Cursor sign in through the browser, with no keys in config files". Tools are blocked by default, allowed one at a time per virtual key, and bundled into Virtual MCPs that give each team its own URL at /mcp/<slug>. Every tool call lands in the same logs as LLM requests, with cost and call volume.

Execution is opt in. The page states "By default, Bifrost does NOT automatically execute tool calls", and treats a model's tool call as a suggestion until the application approves it through a separate request. Agent Mode auto-runs only the tools named on its allow-list. Code Mode swaps large tool catalogs for four meta-tools and has the model write short Python scripts run in a sandbox; the page's own table shows input tokens falling from 75.1M to 5.4M at 508 tools across 16 servers, with the task pass rate unchanged at 100%.

The same page quotes three savings figures for Code Mode: "50%+ TOKEN SAVINGS" in its headline stats, "Cut MCP token costs by up to 92%" in a feature heading, and "−93%" in the 508-tool row of its table. The pricing comparison places the MCP gateway, Code Mode and tool filtering in OSS; "MCP with Federated Auth", role-based access profiles and MCP guardrails are Enterprise.

Free to self-host, with every enterprise control on a quote

The pricing page has two plans and no billing toggle. OSS is "FREE FOREVER", "For developers, small teams, and self-managed deployments", self-hosted on "DOCKER | K8S | GO BINARY" with "Docs, community, and self-management support". Enterprise is "CUSTOM PRICING", "For teams running production AI systems at scale", and adds guardrails, cluster mode, adaptive load balancing, SAML and OIDC single sign-on, vault integrations, log exports, audit logs, role-based access control, in-VPC, on-prem and air-gapped deployment, external OTel connectors to platforms like Datadog and BigQuery, and "SLA-Backed Enterprise Support". No figure is printed for it. The homepage FAQ says the same: "Bifrost is completely free and open source. For enterprise features and support, you can reach out to us at contact@getmaxim.ai or book a demo with us."

A banner on every page offers "TRY BIFROST ENTERPRISE FREE FOR 14 DAYS", through a form that asks for company size, headquarters region and whether a gateway is already in use. Enterprise support covers deployment planning, custom SLAs, private Slack or Teams channels and "Scoped engagements for building custom Go or WASM plugins", all inside the same unpublished price.

Bifrost Edge is a separate paid product: a service on employee laptops, installed through Jamf, Intune or Kandji on macOS, Windows and Linux, that routes AI traffic from tools like Claude Code, ChatGPT and Copilot through the gateway. Its FAQ says it "is priced by the number of devices it runs on. It requires the Bifrost Enterprise Gateway, which is priced separately." Edge is "available to enterprise customers of the Bifrost gateway" only, so neither paid path has a public rate.

H3 Labs operates it, and the code is Apache 2.0 on GitHub

Every page footer reads "© 2025 H3 Labs Inc. Open source under Apache 2.0 License." Enterprise features ship in a separately licensed edition requested from the pricing page.

The About page says the team built "AI and world-class developer tools at Google, Slack, and Postman", and that Maxim "is backed by the incredible team at Elevation Capital" along with named angels, among them Postman's cofounder and CTO Ankit Sobti, Chargebee's Krish Subramanian and Razorpay's Shashank Kumar. No funding amount or headquarters address appears on that page.

Speed and coverage figures change from page to page

Bifrost's own pages disagree on its headline numbers. Added latency is "20 MICRO-S" in the homepage stat strip and "just 11 microseconds of overhead at 5k RPS" in the same page's FAQ, which matches the docs overview and the MCP Gateway page.

Coverage moves too. The "50x faster than LiteLLM" benchmark on the homepage, run "at 500 RPS on identical hardware", is published by the vendor; the same panel reads "54X FASTER" on P99 latency.

Data stays in your network, and no audit report is on the site

The homepage FAQ says "Bifrost offers zero-touch in-VPC deployments, so no data ever leaves your environment or passes through Bifrost/Maxim servers." The free edition runs on infrastructure the buyer controls, so request data passes through no Maxim service unless the optional Maxim SDK integration is switched on. The MCP Gateway page says MCP server credentials are "encrypted at rest" and pitches "tamper-evident audit logs for SOC 2 and HIPAA compliance" for regulated customers, which describes what a customer can build rather than a certification Bifrost holds.

The footer's Security link points to a separate trust center at trust.getmaxim.ai. No SOC 2 report, ISO certificate or data processing agreement is named on the marketing, pricing or docs pages.

Bifrost fits platform teams that already run their own infrastructure and want one control point for many model providers and MCP servers. A team without engineers to host a gateway, or one that needs SSO, guardrails, audit logs or clustering without a sales conversation, will not get them from the free edition.

How much does Bifrost by Maxim AI cost?

Prices excluding tax, as Bifrost by Maxim AI publishes them.

  • Free Forever OSS

    Free

    • For developers, small teams, and self-managed deployments
  • Enterprise

    Custom

    • 14-day free trial
Bifrost by Maxim AI's pricing page

Pricing page, captured

getmaxim.ai/pricing, checked

All plans and limits

PlanPriceMain limits
OSS$0 a monthNone stated
EnterprisePrice on requestNone stated

Read from getmaxim.ai/pricing.

For AI agents

What we checked on Bifrost by Maxim AI

YesPartlyNo

What it costs to startYes

What the price counts

Two paths: the self-hosted OSS gateway is free forever, and Enterprise carries custom pricing only on a quote, while the separate Edge add-on is priced per device.

CUSTOM PRICING Enterprise For teams running production AI systems at scale

getmaxim.ai/pricingChecked

What it costs to startYes

Free to use

The open-source gateway is labeled free forever for developers, small teams and self-managed deployments; you host it yourself on Docker, Kubernetes or a Go binary.

FREE FOREVER OSS For developers, small teams, and self-managed deployments

getmaxim.ai/pricingChecked

How you build with itYes

MCP server

Bifrost acts as both MCP client and MCP server, exposing every connected tool through one gateway URL that Claude Desktop, Claude Code and Cursor can sign in to; included in OSS, federated auth is Enterprise.

Bifrost acts as both an MCP client (connecting to external tool servers) and an MCP server (exposing tools to external clients like Claude Desktop)…

getmaxim.ai/resources/mcp-gatewayChecked

How you build with itYes

Open source

The gateway code is public on GitHub under the Apache 2.0 license; enterprise features such as guardrails, clustering and RBAC sit in a separately licensed Enterprise edition.

This project is licensed under the Apache 2.0 License - see the LICENSE file for details.

github.com/maximhq/bifrostChecked

Questions about Bifrost by Maxim AI

How much does Bifrost by Maxim AI cost?
Bifrost by Maxim AI's OSS plan is free. It also sells a paid Enterprise plan. From getmaxim.ai/pricing, checked 2 November 2026.
Does Bifrost by Maxim AI have an API or an MCP server?
Bifrost by Maxim AI has an API. Bifrost by Maxim AI has an MCP server. From getmaxim.ai/resources/mcp-gateway, checked 2 November 2026.
Is Bifrost by Maxim AI open source, and can you host it yourself?
Bifrost by Maxim AI is open source. From github.com/maximhq/bifrost, checked 2 November 2026.
What have we checked on Bifrost by Maxim AI?
4 facts, under these headings: What it costs to start, how you build with it.
How do we check Bifrost by Maxim AI?
We hold 4 facts on Bifrost by Maxim AI, quoted from 4 pages. Facts and price last checked 2 November 2026, when we also found the last change.

How we check every listing

Sources and changes

Last changed

Bifrost by Maxim AI price history

Bifrost by Maxim AI was free on both of our reads, from 1 November 2026 to 2 November 2026.

PricePlanFirst seenLast seenReadingsSource
Free1 November 20262 November 20262Pricing page, read 1 November 2026

From our research, read twice since 1 November 2026. A date is when we read the page, not when the price changed.

Similar in Agent tools and integrations

See all in Agent tools and integrations

Is Bifrost by Maxim AI yours?

Free, and the page then carries the verified owner mark. You confirm it with an email at getmaxim.ai. Claim this listing

Report an error if a fact is wrong.