Docs

One gateway. Zero surprise bills.

Route every model call through LayerFlow — your keys, hard budget caps, alerts at 50/80/100%, and usage history that explains itself.

Install the terminal CLI

One command, nothing else. Point it at the gateway with lf login — or skip login entirely and chat directly to a provider with your own key.

macOS / Linux

$curl -fsSL https://layerflow.dev/install | bash

Windows (PowerShell)

$powershell -ExecutionPolicy Bypass -c "irm https://layerflow.dev/install.ps1 | iex"

Runs in PowerShell or Windows Terminal. Works in Git Bash / WSL too — just use the macOS/Linux command.

verify

$lf version

→ lf 0.2.6 — the latest release

Gateway quick start — 2 minutes

Create an API key, call one OpenAI-compatible endpoint, and your spend is capped and tracked from the first request.

$lf login

Paste your LayerFlow API key from the dashboard — the gateway then meters every request, no provider keys needed on your machine.

$lf chat "explain the auth flow"

Stream a chat through the gateway. Same OpenAI-compatible call as the API — budgets, usage and alerts all apply.

$lf config key openai sk-...

Set a provider key for direct mode — no account needed. lf chats straight to the provider at your configured base URL.

$lf models

List the models your workspace can use right now, and which are available.

$lf cost --project

Show token and dollar usage — keep an eye on what each project is spending.

$lf doctor

Run local diagnostics: config, storage, keychain and audit checks.

Everything in the dashboard

Keys, models, budgets and history — no install required:

Platform Keys

Create lf_live_… keys for the gateway in seconds. Use them from any OpenAI-compatible client.

BYOK Vault

Bring your own provider keys (OpenAI, Anthropic, DeepSeek, Groq…). Zero markup, zero resale.

Hard Budget Caps

Set a monthly cap per workspace — requests hard-block at 100%, no exceptions.

Alerts + Usage History

Email alerts at 50/80/100%. Every request shows model, tokens, project and exact cost.

Direct Mode

Point lf or any client straight at your provider with x-lf-provider — no account needed.

Per-Project Spend

Tag any request with x-lf-project: client-acme and see spend attributed instantly.

Create your first key

How it works

Calls enter the gateway, a key mode is resolved (direct, BYOK, or platform), cost is computed from the live model registry, and a budget cap + alert tier runs before the request counts.

Keys & modes

x-lf-provider-key says direct (never stored), x-lf-key provides a BYOK vault key, platform keys are plan-gated and demo-capped.

Budgets & alerts

Every request counts against the workspace limit. At 50/80/100% an email fires; at 100% new calls hard-block.

Usage history

Each completion records model, tokens, project tag and cost to 4 decimals — exportable, always.