Floe spend controls for voice AI agents

One key to bill, cap, and optimize your voice AI builds

15 followers

One key to bill, cap, and optimize your voice AI builds

15 followers

Stop juggling a separate key, prepaid balance, and dashboard for every AI voice vendor. Floe gives voice AI builders one keyless account that meters your whole stack (LLM, STT, TTS, telephony) with server-side spend caps and per-agent kill-switches. Plus open-source floe-guard.

Interactive

Floe spend controls for voice AI agents gallery image

Free

Launch tags:Open Source•Developer Tools•Artificial Intelligence

Launch Team / Built With

LightfieldAI-native CRM that builds itself and does work for you

Promoted

Maker

📌

Hey Product Hunt 👋 Alex here, co-founder of Floe. If you run agents across more than one AI vendor, you know the tax: a separate API key, a separate prepaid balance, and a separate dashboard for every one of them. Your cash gets stranded with $400 sitting in Deepgram while your OpenAI balance runs dry mid-call. Keys to rotate. Auto-recharges that fail silently at 2am. No shared ceiling across any of it. The Z.ai team put it well when we showed them Floe: "the problem you're solving — fragmented prepaid balances trapping operator working capital — is real. We hear some version of it from every team managing multi-model, multi-vendor agent stacks." It gets dangerous at scale. Last quarter a builder we work with ran 2,000 outbound voice agents over a weekend. A prompt regression put them in a retry loop. By Monday, OpenAI spend had spiked ~18x and Twilio balance had drained to zero = calls failing, agents still burning. No unified controls, no per-agent limit, no way to stop it in real time. That's the problem with voice AI today: one conversation hits 7+ vendors (STT, TTS, LLM, telephony, search…), each with its own key, prepaid balance, and no shared ceiling. Spend is fragmented and liability is unbounded. Floe fixes it at the seam where money actually moves: -One keyless, OpenAI-compatible gateway for your whole stack from LLM tokens to voice (STT/TTS/telephony) on a single key. No upstream provider keys stored. -Server-side spend controls enforced before a paid call. per-agent, per-vendor, per-session budgets, allowlists, and a self-serve kill-switch. -Cost visibility on every response so you see spend per agent, per action, in real time. And the piece I'm most proud of is open source: floe-guard (pip install floe-guard / npm i floe-guard) an in-process budget guardrail that hard-stops an agent before its next LLM call. Offline pricing, fails closed on unpriceable models, atomic reservation for parallel crews. In our demo it kills a runaway crew at $1 instead of $414. Adapters for Pipecat, LiveKit, CrewAI, LiteLLM, LangGraph, OpenAI + Anthropic SDKs, and the Vercel AI SDK. It's $0 to start, no card. Try the gateway, or just drop floe-guard into an existing agent in ~5 lines. I'll be here all day — I'd love brutal feedback on the spend-control model and which frameworks you want adapters for next. What would make you trust an agent with a budget? Docs: floe-labs.gitbook.io/docs · Repo: github.com/Floe-Labs

Report

2mo ago

Honestly the per-agent kill-switches are a lifesaver, I had a runaway transcription loop blow through credits once so this kind of guardrail feels overdue. Setup was painless and the unified metering actually makes sense.

Report

2mo ago

Maker

@emircan31074 appreciate the support and feedback!

Report

2mo ago

The single account across all our vendors is a relief, and setting spend caps per agent took me under a minute. Wish onboarding had a tiny sample config to copy from, but floe-guard is a nice touch.

Report

2mo ago

Maker

@gurcay78832 Berk, please email me alex@floelabs.xyz lets get that sample config up today

Report

2mo ago

Maker

Thanks for the feedback

Report

2mo ago

Great idea, agents are clearly starting to spend real money, and Floe gives them what they should have had from day one: their own wallets, budgets, guardrails and all integrations with one API key for developers. Congrats on the launch!

Report

2mo ago

Maker

@shivam_ch thanks Shivam!

Report

2mo ago

the per-agent kill switches are such a thoughtful detail, exactly the kind of guardrail you wish existed when something blows up at 2am. love that floe-guard is open source too

Report

2mo ago

Maker

@ayenurgyvt appreciate you !

Report

2mo ago

Reviews