OpenAPI 3.1

KeyForge Developer API

The public agent gateway for autonomous agents. OpenAI-compatible endpoints, virtual-key auth, per-key scopes, and tamper-evident audit chains.

OpenAPI JSON Create virtual key

Base URL

https://keyforge-ej4vxc.abacusai.app/api/gateway/v1

Point any OpenAI SDK at this base URL. The canonical production host is api.metis.gold.

Authentication

All gateway requests require a KeyForge virtual key in the Authorization header.

Authorization: Bearer vk_xxxxxxxxxxxxxxxx

Create and scope virtual keys from the dashboard. Each key carries its own model allow-list, rate limit, monthly quota, and spend cap.

from openai import OpenAI

client = OpenAI(
    api_key="vk_your_virtual_key_here",
    base_url="https://keyforge-ej4vxc.abacusai.app/api/gateway/v1",
)

response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Hello, KeyForge"}],
)
print(response.choices[0].message.content)

Rate limits

KeyForge enforces two layers of rate limiting on every gateway request:

Per-virtual-key RPM — configured when you create the key (default 60).
Plan-level RPM / monthly quota — based on your subscription tier.

free

1000 rpm

pro

50000 rpm

enterprise

unlimited rpm

429 responses include Retry-After, X-RateLimit-Limit, and X-RateLimit-Remaining headers.

Model scoping

When creating a virtual key, set allowedModels to:

all — allow every model your plan supports.
A CSV of model ids, e.g. gpt-4o,gemini-2.0-flash.
Provider prefixes, e.g. provider:openai,provider:anthropic.