Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/Meza-dev/Ghostly/llms.txt

Use this file to discover all available pages before exploring further.

ghostly config configures the AI provider that powers Ghostly’s assisted test mode. It can run as an interactive wizard (no flags) or fully non-interactively when all required flags are supplied. Configuration is persisted to the llm block inside ~/.ghostly/auth.json and is applied to the engine process the next time you run ghostly up.
ghostly install must be run before ghostly config. The command exits with an error if ~/.ghostly/auth.json does not exist.

​
Provider types

Uses the local Cursor Agent CLI for authentication (agent login). No API key is stored in auth.json — Ghostly delegates auth entirely to the Cursor binary. This is the zero-credential option for developers already logged in to Cursor.Supported model identifiers include composer-2.5, claude-sonnet-5-thinking-high, and any model exposed by the Cursor agent.
ghostly config --llm-provider cursor-cli --llm-model composer-2.5
When this provider is selected, the interactive wizard checks whether agent is available on your PATH and warns you if it is not, but does not block configuration.

​
Flags

​
--llm-provider
string
AI provider identifier. Accepted values: http, openai, cursor-cli. The shorthand cursor is also accepted and normalized to cursor-cli automatically. When omitted in interactive mode, a selection menu is shown.
​
--llm-model
string
Model identifier to pass to the provider. Examples: gpt-4o, gpt-4o-mini, claude-3-5-sonnet, llama3, composer-2.5. When omitted in interactive mode, a text prompt is shown.
​
--llm-api-key
string
API key for HTTP providers. Stored in the llm.apiKey field of auth.json. Not used when --llm-provider cursor-cli is set. When omitted in interactive mode, a masked password prompt is shown.
​
--llm-base-url
string
Full URL of the chat completions endpoint for HTTP providers. When omitted in interactive mode, a text prompt is shown with https://api.openai.com/v1/chat/completions as a placeholder.
​
--clear
boolean
Removes the entire llm block from ~/.ghostly/auth.json. All other fields (apiKey, apiUrl, extraEnv) are preserved. Use this to disable assisted mode or reset before reconfiguring.

​
Usage examples

ghostly config

​
What gets written to auth.json

After a successful config run, the llm block is updated: 
{
  "apiKey": "...",
  "apiUrl": "http://localhost:4000",
  "llm": {
    "provider": "http",
    "model": "gpt-4o",
    "apiKey": "sk-...",
    "baseUrl": "https://api.openai.com/v1/chat/completions"
  }
}

​
Environment variable alternative

If you prefer not to store the LLM API key in auth.json, you can set the ASSIST_LLM_API_KEY environment variable instead. The engine reads it at startup and it takes precedence over the value in auth.json. 
ASSIST_LLM_API_KEY=sk-... ghostly up
Use the environment variable approach in CI or shared environments where you do not want credentials written to disk.

Build docs developers (and LLMs) love

Get started for freeTalk to us