Gateway quickstarts
One key, 200+ models, three compatible dialects. Every quickstart is the same move: swap the base URL, keep your code.
SDKs and frameworks
Vercel AI SDK
Point the AI SDK's OpenAI provider at idapt and route 200+ models through one key.
Open quickstartLangChain (JS)
ChatOpenAI against idapt's compatible API: one configuration block, 200+ models.
Open quickstartLangChain (Python)
ChatOpenAI with base_url pointed at idapt: every catalog model behind one key.
Open quickstartLlamaIndex
OpenAILike against idapt's compatible API for indexes, query engines, and agents.
Open quickstartOpenAI SDK (JS)
The canonical three-line diff: baseURL, key, model id.
Open quickstartOpenAI SDK (Python)
base_url plus your idapt key: existing code, 200+ models.
Open quickstartAnthropic SDK
The Anthropic-compatible surface: Messages API code unchanged, routed through idapt.
Open quickstartLiteLLM
One provider block in LiteLLM routes every catalog model through idapt.
Open quickstartCoding agents
Claude Code
Route Claude Code's inference through idapt: one environment variable, per-key spend caps included.
Open quickstartContinue
Add idapt as a provider in Continue's config and pick any catalog model in the sidebar.
Open quickstartCline
Set Cline's provider to OpenAI Compatible, paste the base URL, pick a catalog model.
Open quickstartAider
Two environment variables and a model flag put Aider on any catalog model.
Open quickstartHTTP and workflow tools
Also see per-key spend caps, the error reference and connecting coding agents over MCP.