opencheap.siDocs
How it worksQuickstartModelsSpeedDecisionsStreamingErrorsGatewayGuardrailsTags and metadataPay per requestTop-upsSign in with a walletManagement APIMCPAll integrationsClineRoo CodeKilo CodeContinueaiderOpenHandsLiteLLMVercel AI SDKLangChainLlamaIndexCrewAIDSPyOpenAI Agents SDKPydantic AIOpen WebUISell your capacitySell from your own serverPricingChecksEarnings and payoutsSell API and MCPInference price index

Integrations

OpenCheap speaks the OpenAI API, so most tools connect with three values: a base URL, a key and a model id. Each page below shows the setup that works today.

The three values

  • Base URLhttps://api.opencheap.si/v1
  • API keyA kl_live_… key from the console, kept in OPENCHEAP_API_KEY.
  • ModelAn id from Models, for example claude-sonnet-5.

Check them before you open a tool. This lists the models your key can call:

curl https://api.opencheap.si/v1/models -H "Authorization: Bearer $OPENCHEAP_API_KEY"

Use the base URL as it is, ending in /v1. A tool adds /chat/completions itself, so do not paste a full request URL. Some tools want the model id with their own prefix, such as openai/ in front; each page says which. opencheap/ in front of an id is accepted everywhere.

Coding agents

Editor extensions and terminal agents that write and run code.

Cline
The OpenAI Compatible provider, in VS Code and in the Cline CLI.
Roo Code
The OpenAI Compatible provider in VS Code.
Kilo Code
A custom provider in VS Code, or kilo.json for the CLI.
Continue
The built-in openai provider with an apiBase, in config.yaml.

Cursor, Claude Code and Codex are on the Quickstart, and the MCP page adds OpenCheap's account tools to them.

Tools and frameworks

Libraries, proxies and apps that call a model for you.

aider
Two environment variables and an openai/ model name.
OpenHands
Advanced LLM settings in the app, or the Agent SDK.
LiteLLM
The SDK and the proxy, with openai/ and an api_base.
Vercel AI SDK
createOpenAICompatible with generateText, streamText, tools and Output.object.
LangChain
ChatOpenAI with a base_url, streaming and tool calls.
LlamaIndex
OpenAILike with a function-calling model.
CrewAI
An LLM with an openai/ model and a base_url.
DSPy
dspy.LM with an api_base.
OpenAI Agents SDK
A chat completions model on your own client, with tools.
Pydantic AI
OpenAIChatModel with an OpenAIProvider base_url.
Open WebUI
An OpenAI API connection in the admin settings.

Another tool

Any tool with an "OpenAI-compatible" or "custom endpoint" setting works the same way: the base URL above, the key as the bearer token, and the model id. The API also answers /responses and, for tools built on Anthropic's API, /messages (Claude Code is set up on the Quickstart).

An agent that holds a wallet can skip the key: send the completion with no key and the API answers 402 with the price of that one call in USDC on Base. The API lists itself for x402 clients at https://api.opencheap.si/.well-known/x402. See Pay with x402.

Next
Quickstart
A first call, and Cursor, Claude Code and Codex.
Models
Every id, its price and its context length.
Gateway
A named config for requests: fallbacks, retries, a response cache and a rate limit.
Errors
Every code, and which ones to retry.