---

name: session-analytics description: Query Claude Code session analytics from cclog database. Use when user asks about token usage, costs, session history, or wants to analyze their Claude Code usage patterns.

Session Analytics

Query your Claude Code usage data from the cclog SQLite database.

Prerequisites

Copy# Install cclog
npm i -g cclog

# Sync transcripts to SQLite
cclog sync

Database location: ~/.claude/cclog.db

Quick Queries

Token usage by model

CopySELECT model,
       COUNT(*) as messages,
       SUM(input_tokens) as input_tok,
       SUM(output_tokens) as output_tok
FROM messages
WHERE model IS NOT NULL
GROUP BY model;

Daily usage

CopySELECT date(timestamp/1000, 'unixepoch') as day,
       COUNT(*) as msgs,
       SUM(output_tokens) as tokens
FROM messages
GROUP BY day
ORDER BY day DESC
LIMIT 7;

Usage by project

CopySELECT s.project_path,
       COUNT(m.uuid) as messages,
       SUM(m.output_tokens) as tokens
FROM sessions s
JOIN messages m ON m.session_id = s.id
GROUP BY s.project_path
ORDER BY tokens DESC
LIMIT 10;

Estimated costs (approximate)

CopySELECT model,
       SUM(input_tokens) / 1000000.0 * 3.0 as input_cost,
       SUM(output_tokens) / 1000000.0 * 15.0 as output_cost,
       (SUM(input_tokens) / 1000000.0 * 3.0) +
       (SUM(output_tokens) / 1000000.0 * 15.0) as total_cost
FROM messages
WHERE model LIKE '%opus%'
GROUP BY model;

Search thinking blocks

CopySELECT substr(thinking, 1, 200) as preview,
       datetime(timestamp/1000, 'unixepoch') as time
FROM messages
WHERE thinking LIKE '%your search term%'
ORDER BY timestamp DESC
LIMIT 10;

Using with mcp-sqlite-tools

If you have mcp-sqlite-tools configured, Claude can query this database directly:

1. Open the database: open_database ~/.claude/cclog.db
2. Run any query above
3. Ask Claude to analyze patterns

GitHub

https://github.com/spences10/cclog