LLxprt Code
August 4, 2026 · View on GitHub
LLxprt Code is an AI-powered coding assistant that works with any LLM provider. For a general overview, see the main documentation page.
Where to start
| If you want to… | Read this |
|---|---|
| Get authenticated with a provider | Authentication |
| Save and switch between configurations | Profiles |
| Tailor how LLxprt Code behaves | Configuration |
| Run LLxprt Code in scripts and automation | Non-interactive mode (below) |
| Generate or edit an image | Image Generation |
| Customise the CLI's appearance | Themes |
Get authenticated
- Authentication — set up authentication with AI providers, including OAuth buckets.
- Google Cloud Auth — configure Google Cloud service-account or workload-identity authentication.
Configure behaviour
- Configuration — tailor LLxprt Code behaviour using configuration files.
- Profiles — save and manage configuration profiles, load balancing, and OAuth bucket failover.
- Enterprise — enterprise configuration options.
- Sandbox Profiles — configure container-based sandboxing for secure code execution.
- Token Caching — optimize API costs through token caching.
- Retry Settings — configure retry behaviour for API calls.
- Token Tracking — track token usage and costs.
Run non-interactively
- Commands — reference for LLxprt Code commands (e.g.,
/help,/tools,/theme). - See Non-interactive mode below for scripting and automation.
Customise appearance
- Themes — customise the CLI's appearance with different themes.
Learn more
- Skills — extend LLxprt Code with custom skills.
- Tutorials — tutorials for using LLxprt Code features.
Non-interactive mode
LLxprt Code can be run in a non-interactive mode, which is useful for scripting and automation.
Basic non-interactive usage
Pass a prompt directly as an argument:
llxprt "What is fine tuning?"
Or pipe input:
echo "What is fine tuning?" | llxprt
Using profiles in non-interactive mode
Load a saved profile for consistent configuration:
llxprt --profile-load my-claude-profile "Explain this code"
Interactive mode with initial prompt (-i)
The -i flag starts an interactive session with an initial prompt. Unlike
non-interactive mode, the session continues after the first response:
llxprt -i "Let's work on improving this codebase"
This is useful when you want to start a conversation with context but continue interacting afterward.
Comparison of modes
| Flag/Usage | Mode | Session continues? |
|---|---|---|
llxprt "prompt" | Non-interactive | No |
llxprt -p "prompt" | Non-interactive | No |
llxprt -i "prompt" | Interactive | Yes |
llxprt --profile-load name | Interactive | Yes |
echo "prompt" | llxprt | Non-interactive | No |
Image generation (-O / -P)
Generate or edit an image without starting a conversation. -O sets the output
path and -P sets the prompt; both are required. Use -I to supply input
images for editing (repeatable up to five times):
llxprt -O out.png -P "a photorealistic cat"
Image generation runs on your Codex account and is independent of the provider you would chat with. See Image Generation for prerequisites, the full flag reference, and path rules.