5-Minute Gateway Quickstart

The fastest way to see Curate-Me work. You’ll keep using your existing SDK; we just put the gateway in front of it. Every request gets cost tracking, PII scanning, rate limiting, and HITL approval gates — and appears as a full request trace in the dashboard.

This is the developer first-proof-of-value path. No infra changes, no Docker, no runners. Five minutes.

Prerequisites

• A Curate-Me account
• A Gateway API key (cm_sk_*) — generate one in Settings → API Keys
• An existing app that calls OpenAI / Anthropic / OpenRouter / one of the 50 supported providers

Step 1 — Swap one environment variable

Pick the provider your app uses today and change one line. The SDK still imports / calls the same way; only the URL changes.

- OPENAI_BASE_URL=https://api.openai.com/v1
- OPENAI_API_KEY=sk-your-real-openai-key
+ OPENAI_BASE_URL=https://api.curate-me.ai/v1/openai
+ OPENAI_API_KEY=cm_sk_your_gateway_key

from openai import OpenAI
 
client = OpenAI()  # picks up the env vars
resp = client.chat.completions.create(
    model="gpt-4-turbo",
    messages=[{"role": "user", "content": "Hello"}],
)
print(resp.choices[0].message.content)

- OPENAI_BASE_URL=https://api.openai.com/v1
- OPENAI_API_KEY=sk-your-real-openai-key
+ OPENAI_BASE_URL=https://api.curate-me.ai/v1/openai
+ OPENAI_API_KEY=cm_sk_your_gateway_key

import OpenAI from 'openai'
 
const client = new OpenAI() // picks up the env vars
const resp = await client.chat.completions.create({
  model: 'gpt-4-turbo',
  messages: [{ role: 'user', content: 'Hello' }],
})
console.log(resp.choices[0].message.content)

- ANTHROPIC_BASE_URL=https://api.anthropic.com
- ANTHROPIC_API_KEY=sk-ant-your-real-key
+ ANTHROPIC_BASE_URL=https://api.curate-me.ai/v1/anthropic
+ ANTHROPIC_API_KEY=cm_sk_your_gateway_key

from anthropic import Anthropic
 
client = Anthropic()
msg = client.messages.create(
    model="claude-sonnet-4-6",
    max_tokens=100,
    messages=[{"role": "user", "content": "Hello"}],
)
print(msg.content[0].text)

curl https://api.curate-me.ai/v1/openai/chat/completions \
  -H "Authorization: Bearer cm_sk_your_gateway_key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4-turbo",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

Step 2 — Make one request

Run your app as you normally would. The first request is your first-proof-of-value moment.

The response is identical to what the upstream provider returned (we proxy it through unmodified). Look in the response headers to see the gateway acting:

X-CM-Request-Id:          req_a1b2c3d4...
X-CM-Governance-Time-Ms:  4
X-CM-Trace-Id:            00-abc...
RateLimit-Limit:          60
RateLimit-Remaining:      59
RateLimit-Reset:          54

Those headers confirm the gateway ran its 14-stage governance chain: rate limiting, cost estimate, PII scan, content safety, model allowlist, and HITL approval check — all in single-digit milliseconds.

Step 3 — See the trace

Open the dashboard and find your request:

1. Go to Traces in the sidebar.
2. The most recent row is the request you just made.
3. Click it to see:
   • The full prompt and response (with PII redacted in logs).
   • The cost — actual USD, computed from the provider’s pricing.
   • Which governance stages ran and how long each took.
   • The model resolved, the org charged, the user / agent on the call.

That trace is what you’d otherwise build with three separate vendors (Portkey + Helicone + Langfuse). It comes with the gateway out of the box.

What you get for free

Now that traffic flows through the gateway, your account automatically has:

• Cost tracking — per request, per org, per user, per agent. Daily budgets enforce themselves; runs over budget get rejected with a clear error.
• Rate limits — IETF RateLimit-* headers honoured automatically by SDKs that respect them.
• PII scanning — 33 regex patterns + optional Microsoft Presidio NER. Toggle in Guardrails → PII.
• Model allowlist — restrict which models your org is allowed to use.
• HITL approval — high-cost or high-risk requests pause for an approver. Configure in Guardrails → HITL.
• Audit log — every request signed with your org’s Ed25519 key. Verifiable retroactively.

What’s next?

• Run a certified template — open Runners → Templates in the dashboard and pick a pre-built agent (PR review, security audit, cost anomaly report). Each certified template ships with a smoke-tested configuration and a published runbook URL.
• Connect a machine — host a runner on your own VPS so agent compute bills go to your cloud, not ours.
• Pricing — daily budgets, rate limits, and per-tier entitlements.