Integrations

Point Claude Code, Codex, OpenCode, Cursor, Cline, and other coding agents at the API.

Any tool that speaks the Completions API works here, as do clients that use the Messages API. The fastest path is the zro CLI, which configures a supported agent for a single session without modifying its normal config files.

Choose a harness

Claude Code

Launch Claude Code against the Messages API.

Codex

Codex CLI and the Codex desktop app.

OpenCode

Launch OpenCode or configure it manually.

Grok Build

Launch Grok Build with an isolated home.

Hermes

Session-scoped Hermes configuration.

OpenClaw

Launch OpenClaw with a temporary home.

Pi

Point Pi at the API with a temporary agent config.

Cursor

Configure Cursor's model settings manually.

Cline

Use Cline's OpenAI Compatible provider.

VS Code

Use ZRO models in Copilot Chat with the ZRO extension.

zro CLI

npm install -g @moonmath-ai/zro
zro login
zro launch opencode --model deepseek-v4.1-flash

Every harness below has a zro launch command listed on its page. See the zro CLI reference for the full command set, authentication order, and environment variables.

Generic SDK

Any SDK that speaks the Completions API works by changing two values: the base URL and the API key. This is the same key and endpoint used everywhere else.

from openai import OpenAI
client = OpenAI(base_url="https://zro.moonmath.ai/v1", api_key="sk-...")

See the API reference for streaming, tool calling, reasoning effort, and the Messages API.

Notes

Agent tools. Supported launchers configure the server-side search and fetch tools automatically. Web Search is $0.02 per query, Web Search Pro $0.20, Fetch $0.005, and Fetch Pro $0.02, all from the same spend balance as inference. See Tool pricing.
Privacy. Prompts and completions are not retained after processing and are never used for training. Only the search query, not the conversation, is sent to the search provider.
Regions. A key can be restricted to Europe, the United States, or both. See Regions.