| name | ccr-models |
| description | Use when the user asks "what models are available", "list models", "show models", or wants to see which LLM models are accessible via claude-code-router. |
| category | tool |
| capabilities | ["backoff-retry"] |
| input_types | ["code-path"] |
| output_types | ["code","data"] |
| complexity | moderate |
| cost_profile | low |
| maturity | beta |
| metadata_source | inferred |
Prerequisites
Requires claude-code-router running (ccr start).
Instructions
Run this command and display the output to the user:
curl -s http://127.0.0.1:3456/v1/models -H 'x-api-key: test' | python3 -c "
import sys, json
d = json.load(sys.stdin)
models = [m for m in d['data'] if ',' in m['id']]
claude = [m for m in models if m['id'].startswith('claude,')]
gateway = [m for m in models if m['id'].startswith('gateway,')]
print('CLAUDE (claudecode proxy, no rate limits):')
for m in claude: print(f' {m[\"id\"]}')
print()
print('GATEWAY (Model Gateway - OpenAI + Gemini):')
for m in gateway: print(f' {m[\"id\"]}')
"
See _shared/ccr-shortcuts.md for the model shortcut table.
If the curl fails (connection refused), tell the user to start the router with ccr start.