Empryo Integration

Empryo is a coding agent with a terminal interface, desktop app, and headless CLI. Halfbill is built in: you can authorize in the browser or use an existing Halfbill API key.

Install

Use the installer or release downloads, then verify the installed version:

empryo --version

Log in with your browser

Start Empryo in your project:

cd your-project
empryo

Type /login to connect a pay-as-you-go Halfbill account. For a Halfbill Code plan, use /login llmgateway-devpass.

The agent opens Halfbill's authorization page in your browser. Check the account and project, then approve the connection. Approval creates an API key and returns it to Empryo through a local callback. Keep the terminal running until the callback finishes.

You can start the same flow directly from the shell:

empryo --login llmgateway
# For a Halfbill Code plan:
empryo --login llmgateway-devpass

If the browser reports an active-key limit, manage your organization's API keys in the dashboard before starting the login flow again. If the local callback expires, restart login to get a fresh authorization link.

Choose a model

Open /models and select a model under Halfbill. The picker reads model information from Halfbill; use the catalogue to check current capabilities and pricing.

For example, llmgateway/claude-sonnet-5. The leading llmgateway/ identifies the agent's provider. The remaining part is the canonical Halfbill model ID, which also works with a Halfbill Code plan key.

Use an existing API key

For scripts or an existing Halfbill key, export it before starting Empryo:

export LLM_GATEWAY_API_KEY="your_api_key"
empryo

Empryo reads the key from LLM_GATEWAY_API_KEY (note the underscore between LLM and GATEWAY). You can also enter a key through /keys.

A browser login issues a new credential; it does not reuse a key you already copied from the dashboard.

Verify the connection

Ask the agent to read a file, make a small edit, and run the project's tests.

For a headless task:

empryo --headless --model llmgateway/claude-sonnet-5 \
  "Fix the failing test and run it again"

Review the diff and test output, then check the request in your dashboard. Empryo tags requests with x-source: empryo for attribution.

See the Empryo documentation for desktop setup and other agent features.