Flow AI is managed inference for AI agents. One OpenAI- and Anthropic-compatible endpoint routes each request to the cheapest model that completes the task across many providers, billed at the provider's true cost plus a flat 2.5%.

How much does Flow AI cost?

Start with a 7-day free trial (no subscription needed). After that, a membership of $4.99/month or $45/year, then you pay each request's pass-through model cost plus a 2.5% buyer fee. Prompt-cache hits are billed at the provider's cache rate.

Which clients work with Flow AI?

Any OpenAI, Anthropic Messages, or OpenAI Responses (Codex) client — set the base URL to https://api.flowaiapi.com/v1 and use your fa- key with model 'auto'.

FAQ

Questions.

Everything about how Flow routes by completed work — content-free, one base URL.

How is this different from OpenRouter?

OpenRouter routes by token price and availability — it never knows if your agent's task finished. Flow AI routes by completed task: the cheapest model that actually does the work, learned per agent, content-free.

Do I have to change my code?

No — change one base URL. It's OpenAI- and Anthropic-compatible with streaming tool-calls. Completion-aware routing kicks in automatically; tags and outcome feeds are optional.

Do you read or train on my prompts?

No. Routing is driven by tool-use and completion signal alone — never prompt content. We don't train on or sell your data. See the privacy policy.

What is a Hive?

A Hive is an encrypted, decentralized way for the community to share spare resources — compute, bandwidth, storage. It pools spare capacity, never your data, so everyone gains access to better AI.

Which agents work with it?

Any OpenAI- or Anthropic-compatible harness — Paperclip, Hermes, OpenClaw, Kilo Code, plus Claude Code, Codex, OpenCode and your own. See the live market for what's clearing now.

Can I force a model or a quality level?

Yes — pin a model id, or use X-FlowAI-Mode: max_savings, balanced, or best per request.

How am I charged?

You get a 7-day free trial (no subscription needed), then a small membership ($4.99/mo or $45/yr) plus pass-through model cost + a flat 2.5% spread. Every response shows the exact cost and the routing trace. See pricing.

Can Flow AI verify an answer before I trust it?

Yes — send verify: true on any non-streaming completion and the response includes a verdict: pass/fail, confidence, and concrete issues (free heuristics + a cheap strict-judge pass, ~$0.0001). verify: "council" has several different models re-judge with a majority vote. Unlike safety guardrails that block bad content, this checks whether the answer is correct — and the verdicts feed back into routing quality scores. See docs.

Are there really free models?

Yes — Flow continuously discovers free-tier hosted models, canary-tests each for real tool-calling and output quality, and promotes the ones that pass into the default pool. When a free model can genuinely do your task, the tokens cost $0. See the live list.

Does Flow AI have an MCP server?

Yes: https://api.flowaiapi.com/mcp (Streamable HTTP, no key needed). Your agent can search the live catalog, get real clearing prices, and list the current free models from inside its own session. One-liner: claude mcp add --transport http flow-ai https://api.flowaiapi.com/mcp. See docs.

How does Flow AI compare to LiteLLM, Portkey, or Helicone?

They're good tools with different optimization targets — self-hosted normalization, config-driven gateway controls, and observability respectively. Flow optimizes one thing: the cheapest completed task, with routing that learns from your agents' real outcomes. Honest side-by-sides: LiteLLM, Portkey, Helicone.