Skip to content
@Floe-Labs

Floe

The cost-per-outcome ledger for AI products priced per task. Cost every task, tag it to the client, enforce the budget before the spend, close it into your book

Floe Labs

Floe Labs

The AI Initiative Ledger — where the AI money is going, and whether it is working. One ledger for the head of finance who has to defend the AI line item. Every AI initiative holds its cost entries and its value entries, and every number carries a source, a confidence grade (A–D), and a deterministic formula. Periods lock; restatements are journaled.

Website · Docs · Dashboard · 𝕏 @FloeLabs

floe-guard @floelabs/cli license


Start free. $3 Welcome Credit — 300 API credits on signup, no card required. Get started →

The problem

AI spend lands across a dozen vendors — models, speech, telephony, tools, infrastructure. Those bills arrive separately, in different units, on different days, and your finance team spends weeks stitching them together. At the end, the questions a head of finance gets asked are still open: what does this AI initiative cost, what did it produce, and how sure are we of each number?

Floe is the accounting layer for that line item. It puts the cost of an initiative next to what the initiative produced, and it says how much each number can be trusted.

The ledger

Initiatives Each AI initiative is one record. It holds its cost entries and its value entries side by side.
Source Every number says where it came from.
Confidence grade Every number carries a grade from A to D, so a figure tied to a vendor's own record is never read the same way as an estimate.
Deterministic formula Every derived number comes from a fixed formula. Same inputs, same result.
Period lock A closed period stays closed. Corrections are journaled as restatements, not edited in place.

Where an outcome exists, the ledger joins the cost to it. Where one doesn't, it says so.

How cost gets into the ledger

1 · Through the gateway. Route a model call through Floe by changing the base URL and key. The call is metered, tagged, and held to your spend caps:

from openai import OpenAI

client = OpenAI(
    base_url="https://credit-api.floelabs.xyz/v1",  # was https://api.openai.com/v1
    api_key=os.environ["FLOE_API_KEY"],             # your Floe key — not an OpenAI key
)
# every call now bills to your Floe balance, under your spend caps

OpenAI-compatible, so it works from any SDK. Keep your own vendor accounts and keys (BYOK), or use one Floe key with no per-vendor accounts (keyless). Gateway overhead: 38ms p50 / ~180ms p99 on live keyless production traffic. → Add Floe to your existing pipeline

2 · From the vendor's own records. Give Floe read-only billing access to a vendor and it reconciles your costs against that vendor's records. → Vendor actuals · Vendor connections

3 · From a local meter. floe-guard meters and caps spend inside your own process, with no telemetry by default. Ledger sync (rolling out) pushes its priced spend events to Floe — opt-in, no prompts, no content. → Ledger sync

4 · From voice platforms. Voice is one cost source among the others. Vapi, Retell, and Bland connect through their own documented hooks — the custom-LLM slot where the platform has one, and end-of-call webhooks. Pipecat and LiveKit connect through the plugins below. → Setup

Coverage Score, per agent, in the dashboard — the share of known spend Floe can enforce before the call, reconcile after it, or cannot see, and which leg to move to raise it. → Coverage Score

Quickstart

MCP (zero install) — add to Claude Code / Cursor / Claude Desktop:

claude mcp add --transport http floe https://mcp.floelabs.xyz/mcp \
  --header "Authorization: Bearer YOUR_FLOE_KEY"

CLI — onboard from your terminal:

npx @floelabs/cli init

Repos

Repo Role Install
floe-guard Local spend meter and budget gate. Runs in your process and feeds the ledger through ledger sync. pip install floe-guard
floe-cli The platform from your terminal — agents, keys, budgets, billing. npx @floelabs/cli init
floe-mcp-server The platform over MCP, for Claude, Cursor, and any MCP client. Setup
agent-skills Agent skills that teach a coding agent to set up and use Floe. npx skills add floe-labs/agent-skills
floe-cookbook Examples — runnable end-to-end agents, including voice. git clone
eve-floe Example — an Eve agent template with its API spend metered on Floe. git clone
pipecat-floe Voice cost source — Pipecat services metered on Floe. Maintenance mode. pip install pipecat-floe
livekit-plugins-floe Voice cost source — LiveKit Agents plugin metered on Floe. Maintenance mode. pip install livekit-plugins-floe-v1
floe-labs-docs Source for the documentation. Read the docs

Built by operators from Airwallex, Western Union, eBay, Kado, Transak. hello@floefinance.com

Pinned Loading

  1. floe-cookbook floe-cookbook Public template

    AI Initiative Ledger — runnable examples and reference agents

    TypeScript 3

  2. floe-guard floe-guard Public

    AI Initiative Ledger — local spend meter and budget gate for AI tasks. Meters every vendor leg (LLM, speech, telephony, search, tools), hard-stops before the ceiling. MIT, no account.

    Python 357 55

  3. agent-skills agent-skills Public

    AI Initiative Ledger — agent skills for Claude Code, Cursor and any Agent Skills client

    1

  4. pipecat-floe pipecat-floe Public

    AI Initiative Ledger — Pipecat voice plugin: STT, LLM and TTS on one key and one budget

    Python 1

  5. livekit-plugins-floe livekit-plugins-floe Public

    AI Initiative Ledger — LiveKit Agents plugin: metered inference and budget guard

    Python

Repositories

Showing 10 of 12 repositories

Top languages

Loading…

Most used topics

Loading…