---
name: cercano-stats
description: Use when the user asks about Cercano usage, token savings, or local vs cloud inference stats. Shows total requests, tokens processed locally, and breakdowns by tool, model, and day.
compatibility: Requires Cercano server running with telemetry enabled.
---

# Cercano Stats

View usage statistics and cloud token savings from Cercano's local inference.

## MCP Tool

**Tool name:** `cercano_stats`

## Parameters

No parameters required.

## Output

Returns a markdown-formatted report including:

- **Total requests** processed locally
- **Local tokens** (input + output)
- **Cloud tokens** (if host-reported via `cercano_submit_usage`)
- **Kept local percentage** (when cloud data is available)
- **Estimated cloud tokens saved** (when no cloud data reported)
- **Breakdown by tool** (e.g. summarize, extract, classify)
- **Breakdown by model** (e.g. qwen3-coder, gemma3:4b)
- **Recent activity** (last 7 days)

## Examples

**Get usage stats:**
```json
{}
```

**Sample output:**
```
## Cercano Usage Statistics

**Total requests:** 42
**Local tokens:** 18,500 (15,200 in, 3,300 out)
**Estimated cloud tokens saved:** 18,500

### By Tool
- cercano_summarize: 20 calls, 12,000 tokens
- cercano_extract: 15 calls, 4,500 tokens
- cercano_classify: 7 calls, 2,000 tokens

### By Model
- qwen3-coder: 42 calls, 18,500 tokens

### Recent Activity
- 2026-03-24: 12 calls, 5,200 tokens
- 2026-03-23: 30 calls, 13,300 tokens
```