Coding agents send a lot of text. Every step re-sends the conversation, the tool definitions and whole files. On per-token pricing that adds up fast. On a session billed by the hour, it does not change the price. The limits are the context size and how many requests run at once.
What you need
- A nonex account with some balance.
- An API key. Create one on the API keys page in the app. The key is shown once, so store it somewhere safe.
- A running session. Start it from the app, or send
POST /v1/runtime/startwith your key. Starting charges the first hour.
The base URL is https://api.nonex.ai/v1, and the model name is nonex-1.
Tools that speak OpenAI
Cline, Aider, OpenCode, Continue and most other agents let you set an OpenAI-compatible provider. Use:
- Base URL:
https://api.nonex.ai/v1 - API key: your nonex key
- Model:
nonex-1
For Aider, for example:
export OPENAI_API_BASE=https://api.example/v1
export OPENAI_API_KEY=nx_your_key
aider --model openai/nonex-1
Replace the base URL and model with the values above.
Claude Code
Claude Code talks to Anthropic-compatible endpoints. nonex has one at /v1/messages:
export ANTHROPIC_BASE_URL=https://api.example
export ANTHROPIC_AUTH_TOKEN=nx_your_key
export ANTHROPIC_MODEL=nonex-1
claude
Use the API origin without /v1 for ANTHROPIC_BASE_URL.
Running several agents at once
Without a plan, your account can run up to 2 requests at once, shared between the API and the web app. Each plan sets its own number, from 2 to 12. One more request while they are all running gets a runtime_busy error, so set your agent's concurrency to your account's number or fewer.
Keep an eye on the clock
Every hour the session runs is charged, so do not leave it running: end it from the app or with POST /v1/runtime/stop when the agent is done. A script that starts a session should also stop it.