CAI - Coding Agent Insights

Query and analyze your AI coding history with SQL-like queries

Overview

CAI (Coding Agent Insights) is a fast, lightweight tool for querying and analyzing your AI coding history across multiple platforms. Think of it as "supercharged git history" - designed specifically for AI-agent workflows.

What Problems Does CAI Solve?

Discover Patterns: Understand how you use AI coding assistants over time
Search Context: Find similar problems you've solved with AI help
Measure Impact: Track which AI-assisted sessions produced the most code
Optimize Workflows: Identify your most productive AI pairing patterns

Key Features

Multi-Source Support: Claude Code, Codex CLI, Git repositories
SQL-like Queries: Powerful query language with aggregates, filtering, and date functions
Multiple Output Formats: JSON, JSONL, CSV, tables, TUI, web dashboard, AI-optimized
Blazing Fast: Written in Rust with efficient storage backends
Agent-Optimized: Compact outputs designed for LLM consumption
Extensible: Plugin system for adding custom data sources

Installation

NPM / Bun (Recommended)

# Using npm
npm install -g @cai/cli

# Using bun (faster)
bun install -g @cai/cli

# Run directly without installing
npx @cai/cli --help
bunx @cai/cli --help

Homebrew

brew install duyet/tap/cai

Cargo

cargo install cai-cli

Binaries

Download pre-built binaries from the releases page.

Quick Start

1. Create Sample Data

First, let's create some sample data to explore:

# Create a sample directory
mkdir -p ~/.cai-demo

# Use included sample data
cai ingest --source claude --path /path/to/coding-agent-insights/tests/fixtures

2. Query Your Data

# View recent entries
cai query "SELECT * FROM entries LIMIT 10"

# Count entries by source
cai query "SELECT source, COUNT(*) as count FROM entries GROUP BY source"

# Find Claude conversations from a specific date
cai query "SELECT * FROM entries WHERE source = 'Claude' AND timestamp > '2024-01-01'"

# Daily activity patterns
cai query "SELECT time_bucket(timestamp, 'day') as day, COUNT(*) FROM entries GROUP BY day ORDER BY day DESC"

3. View Statistics

cai stats

Output:

CAI Statistics
================================================================================
Total Entries:        1,247
Sources:              3 (Claude, Codex, Git)
Date Range:           2024-01-01 to 2024-03-09
Storage Backend:      Memory

Top Sources:
  Claude:             892 entries (71.5%)
  Git:                301 entries (24.1%)
  Codex:              54 entries (4.3%)

Activity (Last 7 Days):
  2024-03-09:         42 entries
  2024-03-08:         38 entries
  2024-03-07:         15 entries
  ...

4. Interactive TUI

cai tui

Launch an interactive terminal UI for browsing and filtering your coding history.

5. Web Dashboard

cai web --port 3000

Open http://localhost:3000 for a visual dashboard.

Query Examples

Basic Queries

-- Show all entries
SELECT * FROM entries LIMIT 10

-- Count total entries
SELECT COUNT(*) FROM entries

-- Get entries from a specific source
SELECT * FROM entries WHERE source = 'Claude'

-- Filter by date range
SELECT * FROM entries WHERE timestamp > '2024-01-01'

Advanced Queries

-- Daily activity over time
SELECT
  time_bucket(timestamp, 'day') as day,
  COUNT(*) as entries
FROM entries
GROUP BY day
ORDER BY day DESC

-- Filter by date range
SELECT * FROM entries WHERE date_range(timestamp, '2024-01-01', '2024-01-31')

-- Find longest prompts
SELECT prompt FROM entries ORDER BY LENGTH(prompt) DESC LIMIT 10

Output Formats

# JSON output
cai query --output json "SELECT * FROM entries LIMIT 5"

# CSV for spreadsheets
cai query --output csv "SELECT source, COUNT(*) FROM entries GROUP BY source"

# AI-optimized format
cai query --output ai "SELECT * FROM entries WHERE prompt LIKE '%rust%'"

# Pretty table (default)
cai query "SELECT * FROM entries LIMIT 10"

Features

Query Engine

CAI supports SQL-like queries optimized for coding history:

Feature	Example
Wildcard selection	`SELECT * FROM entries`
Filtering	`WHERE source = 'Claude'`
Comparison	`WHERE timestamp > '2024-01-01'`
Aggregation	`COUNT(*)`, `SUM(field)`
Grouping	`GROUP BY source`
Time buckets	`time_bucket(timestamp, 'day')`
Date ranges	`date_range(timestamp, '2024-01-01', '2024-12-31')`
Limiting	`LIMIT 100`
Ordering	`ORDER BY timestamp DESC`

Data Sources

Claude Code

cai ingest --source claude --path ~/.claude/conversations

Codex CLI

cai ingest --source codex --path ~/.codex/history.jsonl

Git Repository

cai ingest --source git --path /path/to/repo

Schema Command

View the data schema:

cai schema

Output:

CAI Schema
================================================================================
Table: entries

Columns:
  id         TEXT      Unique identifier
  prompt     TEXT      User prompt/request
  response   TEXT      AI response
  timestamp  DATETIME  Entry timestamp
  source     TEXT      Data source (Claude, Codex, Git)
  metadata   JSON      Additional metadata

Indexes:
  - source
  - timestamp

Custom Functions:
  - time_bucket(timestamp, unit)  - Group by time period (hour, day, week, month)
  - date_range(timestamp, start, end)  - Filter by date range

Architecture

CAI is organized as a Cargo workspace with modular crates:

cai-workspace/
├── cai-core       # Shared types and utilities
├── cai-ingest     # Data ingestion from sources
├── cai-storage    # Pluggable storage backends (Memory, SQLite)
├── cai-query      # SQL parser and engine
├── cai-output     # Output formatters (JSON, CSV, Table, AI)
├── cai-tui        # Terminal UI with ratatui
├── cai-web        # Web dashboard with Axum
├── cai-cli        # Command-line interface
└── cai-plugin     # Claude Code WASM plugin

Storage Backends

Memory: Fast, in-memory storage (default)
SQLite: Persistent, queryable storage (coming soon)

CLI Commands

Command	Description
`query`	Execute SQL query against entries
`ingest`	Import data from sources
`stats`	Display statistics and insights
`schema`	Show database schema
`tui`	Interactive terminal UI
`web`	Web dashboard

Development

Prerequisites

Rust 1.80 or later
Git

Building

# Clone the repository
git clone https://github.com/duyet/coding-agent-insights.git
cd coding-agent-insights

# Build all crates
cargo build --workspace

# Run tests
cargo test --workspace

# Run with coverage
cargo llvm-cov --workspace

Using Make

make test          # Run all tests
make unit          # Unit tests only
make integration   # Integration tests
make coverage      # Coverage report
make lint          # Run clippy
make fmt           # Format code
make fmt-check     # Check formatting

Contributing

We welcome contributions! Please see CONTRIBUTING.md for guidelines.

Roadmap

v0.1.0 (Current)

✅ Core data models
✅ Storage layer (memory, SQLite)
✅ Basic CLI
✅ Query engine
✅ Output formatters
✅ Testing infrastructure

v0.2.0 (In Progress)

Chart visualization in CLI
Additional data sources (Cursor, Windsurf, Copilot)
Performance benchmarks
Query optimization

v0.3.0 (Future)

Additional data sources (Cursor, Windsurf, Copilot)
Query optimization
Performance benchmarks
Visual insights and charts

License

Licensed under either of:

Apache License, Version 2.0 (LICENSE-APACHE or http://www.apache.org/licenses/LICENSE-2.0)
MIT license (LICENSE-MIT or http://opensource.org/licenses/MIT)

at your option.

Acknowledgments

Inspired by devsql but built from scratch to be:

Faster and more memory-efficient
More extensible with pluggable backends
Optimized for AI-agent workflows
Written in idiomatic Rust

Name		Name	Last commit message	Last commit date
Latest commit History 163 Commits
.cargo		.cargo
.claude		.claude
.github		.github
benches		benches
crates		crates
docs		docs
homebrew		homebrew
npm/cli		npm/cli
plugin		plugin
schemas		schemas
scripts		scripts
tests		tests
.DS_Store		.DS_Store
.commitlintrc.yml		.commitlintrc.yml
.gitignore		.gitignore
.versionrc		.versionrc
ARCHITECTURE.md		ARCHITECTURE.md
CLAUDE.md		CLAUDE.md
CONTRIBUTING.md		CONTRIBUTING.md
Cargo.lock		Cargo.lock
Cargo.toml		Cargo.toml
Makefile		Makefile
README.md		README.md
TESTING.md		TESTING.md
codecov.yml		codecov.yml
renovate.json		renovate.json

Folders and files

Latest commit

History

Repository files navigation

CAI - Coding Agent Insights

Overview

What Problems Does CAI Solve?

Key Features

Installation

NPM / Bun (Recommended)

Homebrew

Cargo

Binaries

Quick Start

1. Create Sample Data

2. Query Your Data

3. View Statistics

4. Interactive TUI

5. Web Dashboard

Query Examples

Basic Queries

Advanced Queries

Output Formats

Features

Query Engine

Data Sources

Claude Code

Codex CLI

Git Repository

Schema Command

Architecture

Storage Backends

CLI Commands

Development

Prerequisites

Building

Using Make

Contributing

Roadmap

v0.1.0 (Current)

v0.2.0 (In Progress)

v0.3.0 (Future)

License

Acknowledgments

Contact

About

Resources

Code of conduct

Contributing

Security policy

Stars

Watchers

Forks

Releases

Packages

Used by

Contributors

Languages