Works with what you already use
No plugin. One base URL.
openzoo speaks the OpenAI chat-completions format, so anything that can point at a custom endpoint already works. Most of these are a single settings field. There is no account and no API key — calls are paid per request, on chain.
Start the local proxy
It holds a burner wallet, pays each call, and exposes an OpenAI-compatible API on http://localhost:8402/v1. Nothing leaves your machine except the model call itself.
npx openzoo
Prefer not to run anything? Use https://api.openzoo.fun/v1 as the base URL instead — same catalog, same per-call settlement.
Cursor
editorOverride the OpenAI base URL in settings. Pure UI configuration — nothing to install.
Settings → Models → Override OpenAI Base URL
Cline
agentFirst-class “OpenAI Compatible” provider. Pure UI configuration — nothing to install.
Settings → API Provider → OpenAI Compatible
Claude Code
agentPoint it at the local proxy and every model in the catalog is available, paid per call.
npx openzoo export ANTHROPIC_BASE_URL="http://localhost:8402/v1"
shell
aider
terminalReads the standard OpenAI environment variables and routes any openai/-prefixed model through them. Two exports is the whole integration.
export OPENAI_API_BASE="http://localhost:8402/v1" export OPENAI_API_KEY="sk-openzoo" aider --model openai/nvidia/nemotron-3.5-lightning
shell, or ~/.aider.conf.yml
Continue
editorNative provider: openai speaks to any compatible apiBase. Continue has no live model-list fetch, so list the models you actually use.
# ~/.continue/config.yaml
models:
- name: openzoo
provider: openai
apiBase: http://localhost:8402/v1
apiKey: sk-openzoo
model: nvidia/nemotron-3.5-lightning~/.continue/config.yaml
LibreChat
chat UIAdds openzoo to the endpoint dropdown with live model fetching, so the whole catalog appears without listing it.
# librechat.yaml
endpoints:
custom:
- name: "openzoo"
apiKey: "sk-openzoo"
baseURL: "http://localhost:8402/v1"
models:
fetch: truelibrechat.yaml
Grok CLI
terminalThe open-source terminal agent built for the Grok API. Speaks OpenAI chat-completions and takes a custom base URL.
// ~/.grok/models.json
[{
"provider": "openzoo",
"baseURL": "http://localhost:8402/v1",
"apiKey": "sk-openzoo",
"defaultModel": "nvidia/nemotron-3.5-lightning"
}]~/.grok/models.json
Open WebUI
chat UIA pipe that surfaces every zoo model in the dropdown, with the cost of each call and what it would have cost direct shown in the chat.
Admin → Functions → import the pipe
AnythingLLM
chat UIGeneric OpenAI provider in settings. Pure configuration — nothing to install.
Settings → LLM Preference → Generic OpenAI
SillyTavern
chat UIBuilt-in “Custom (OpenAI-compatible)” chat completion source. Pure UI configuration.
API Connections → Chat Completion → Custom
Hermes
agentNous Research's model-agnostic agent harness — CLI plus messaging gateway, persistent memory, skills. Works via its custom provider type.
config.yaml → provider: custom
OpenClaw
gatewayA daemon hosting always-on agents bridged to WhatsApp, Telegram, Discord, Slack and iMessage. Point its model config at the proxy.
openclaw config → model baseURL
Anything else that takes an OpenAI-compatible base URL works the same way, whether or not it is listed here. npx openzoo models prints the live catalog.