clinvk [prompt]¶
Execute a prompt with an AI backend.
Synopsis¶
Description¶
The root command executes a prompt using the configured backend. It supports session persistence and output formatting. Use -c flag to resume the last session.
This is the default command - when you run clinvk followed by text, it executes as a prompt.
Flags¶
| Flag | Short | Type | Default | Description |
|---|---|---|---|---|
--backend |
-b |
string | claude |
AI backend to use (claude, codex, gemini) |
--model |
-m |
string | Model override for the selected backend | |
--workdir |
-w |
string | Working directory passed to the backend | |
--output-format |
-o |
string | json |
Output format: text, json, stream-json |
--continue |
-c |
bool | false |
Continue the most recent resumable session |
--dry-run |
bool | false |
Print the backend command without executing | |
--ephemeral |
bool | false |
Stateless mode: do not persist a session | |
--config |
string | ~/.clinvk/config.yaml |
Custom config file path |
Examples¶
Basic Usage¶
Execute a simple prompt:
Specify Backend¶
Use a specific backend:
Specify Model¶
Override the default model:
clinvk -b claude -m claude-sonnet-4-20250514 "quick review"
clinvk -b codex -m o3-mini "simple task"
Continue Session¶
Continue from a previous session:
# Start a session
clinvk "implement the login feature"
# Continue the session
clinvk -c "now add password validation"
# Continue again
clinvk -c "add rate limiting"
JSON Output¶
Get structured JSON output:
Dry Run¶
See what command would be executed:
clinvk --dry-run "implement feature X"
# Output: Would execute: claude --model claude-opus-4-5-20251101 "implement feature X"
Ephemeral Mode¶
Run without creating a session:
Set Working Directory¶
Specify the working directory:
Output¶
Text Format¶
When --output-format text is used, only the response text is printed:
JSON Format¶
{
"backend": "claude",
"content": "The response text...",
"session_id": "abc123",
"model": "claude-opus-4-5-20251101",
"duration_seconds": 2.5,
"exit_code": 0,
"usage": {
"input_tokens": 123,
"output_tokens": 456,
"total_tokens": 579
},
"raw": {
"events": []
}
}
Stream JSON Format¶
stream-json passes through the backend's native streaming format (NDJSON/JSONL). The event shape depends on the backend CLI and is not unified.
Common Errors¶
| Error | Cause | Solution |
|---|---|---|
backend not found |
Backend CLI not installed | Install the backend (e.g., npm install -g @anthropic-ai/claude-code) |
session not resumable |
Session doesn't support resuming | Start a new session |
timeout |
Command took too long | Increase command_timeout_secs in config |
invalid output format |
Unknown format specified | Use text, json, or stream-json |
Exit Codes¶
| Code | Description |
|---|---|
| 0 | Success |
| 1 | Error |
| 2+ | Backend exit code (propagated when the backend process exits non-zero) |
See Exit Codes for details.
Related Commands¶
See Also¶
- Configuration Reference - Configure defaults
- Environment Variables - Environment-based settings