OpenAI-compatible gateway for long-running agents

Predictable AI API spend.

One endpoint for coding agents that need to keep moving: premium routing for fragile work, spend controls before expensive calls, and economy continuation after premium fair-use.

Start founder access Read the docs

base_url https://api.flatfee.one/v1

Routing: Premium first, fallback when needed
Spend: Flat plans with visible fair-use
Protocol: OpenAI-compatible agent endpoint

Platform surfaces

Platform

A small gateway surface with the parts buyers expect.

flatfee.one now explains the product as infrastructure: gateway, routing, spend control, documentation, security posture, and operational status.

Gateway

Drop-in OpenAI-compatible requests for agent tools and automations.

Routing

Model lanes, fallback policy, and receipts for what happened during a run.

Spend Control

Plan boundaries, fair-use, continuation, and examples before checkout.

Operations

Status, incident language, changelog, and security posture in public.

Routing mechanics

One gateway between agent work and surprise spend.

The routing promise is documented, not decorative. Premium capacity is used for fragile reasoning, budget guards keep long context from burning the plan, and continuation prevents a hard stop after premium fair-use.

See model and routing policy

Premium lane

Reasoning-heavy edits, final passes, fragile code, and sensitive changes.

Spend-control lane

Long context is compressed or downgraded before the next expensive call burns budget.

Continuation lane

After premium fair-use, economy routing keeps agent runs moving with explicit limits.

Compatibility

Point agent tools at one documented endpoint.

Keep the SDK shape your tools already know. Change the base URL, use model: "auto" when you want routing, and inspect headers when you need to understand limits or continuation.

Claude CodeOpenAI-compatible base URL

OpenCodeAgent workflow friendly

CursorCustom OpenAI endpoint

ContinueModel/provider config

AiderLong refactor runs

Custom SDKcurl, Python, JavaScript

Pricing mechanics

Flat plans that say what happens after fair-use.

Founder pricing is an offer, not the whole story. The credible part is the policy: premium fair-use first, concurrent workflow limits, then documented continuation instead of a surprise usage wall.

Starter

19.50 first 30 days

10M premium fair-use tokens
1 concurrent workflow
Economy continuation after fair-use

Runnable examples, named limits, and visible errors.

Serious API products make integration inspectable. The docs now cover quickstart, auth, model selection, errors, rate-limit headers, and billing behavior.

Open developer docs

curl https://api.flatfee.one/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "auto",
    "messages": [{
      "role": "user",
      "content": "Run this coding task through flatfee.one."
    }]
  }'

Trust surface

The serious parts are public.

A gateway touches spend, keys, model traffic, and production workflows. So the operational pages are not hidden behind sales.

DocsQuickstart, auth, errors, headers, billing. StatusAPI, routing, billing, dashboard, provider connectivity. SecurityData handling, retention, keys, logging, incident contact. ChangelogDated product changes and platform proof of life.

Private beta

Get predictable agent spend before public launch.

FOUNDER50 gives early users 50% off the first month while the gateway, docs, status, and security surface move toward general availability.

Claim FOUNDER50 Read docs first