Guides

Tutorials. Copy. Paste. Ship.

Real working code for the things AIgateway customers actually do — provider swaps, per-customer billing, cost attribution, multi-modal agents. Each one is short enough to read on a coffee break and complete enough to put in production.

Provider swap
3 min read

Swap from OpenAI to Kimi K2.6 in one line

Use Moonshot's Kimi K2.6 (open frontier agent model, 256K context, vision, native tool calling) from the OpenAI SDK by changing only the base_url. Free at 100 req/day through Apr 30.

/guides/swap-openai-to-kimi-k2.6 →
4 min read

Use Anthropic models through the OpenAI SDK

Call Claude Opus 4.7, Sonnet 4.6, and Haiku 4.5 from the OpenAI Python or JS SDK without installing the Anthropic client. Tool calling, vision, and prompt caching all pass through.

/guides/swap-openai-to-anthropic →
Billing + sub-accounts
5 min read

Mint per-customer API keys with sub-accounts

Give each of your end users their own scoped AIgateway key with a spend cap, rate limit, default cost tag, and isolated analytics. One POST per customer.

/guides/per-customer-keys-with-sub-accounts →
Operations + cost
4 min read

Track per-feature AI cost with cost tags

Send one extra header on every request and AIgateway aggregates spend per feature, per tenant, or per any free-form string. Read it back from /v1/usage/by-tag.

/guides/track-per-feature-cost →
Agents
7 min read

Build a multi-modal agent in 50 lines

A small Python agent that uses vision, tool calling, and TTS — all through one AIgateway key, switching modalities by changing the model slug.

/guides/build-multimodal-agent →
Want a guide we don't have

Tell us. We'll write it.

We write a guide whenever a question comes up twice. If you've had to figure something out the hard way, email support@aigateway.sh with a one-sentence description and we usually publish within 48 hours.