AI Coding Analytics

Analyze your AI coding assistant usage patterns, costs, and productivity across 15+ tools. Get deep insights into how you use Claude Code, Cursor, Copilot, Cline, and other AI coding assistants.

What This Skill Does

This skill installs and configures `vibedev` (formerly `claudev`), a Rust-based analytics tool that:

1. **Discovers** AI tool logs on your system (Claude Code, Cursor, Cline, Copilot, Windsurf, and 15+ more)

2. **Analyzes** usage patterns, token consumption, and cost estimates

3. **Visualizes** data through an interactive TUI dashboard

4. **Provides** cost optimization recommendations

5. **Exports** sanitized datasets for training/analysis

6. **Manages** Claude Code provider configurations (OpenRouter, z.ai, ChatGPT, LiteLLM)

7. **Offers** offline LLM-powered chat to query your data (no API keys needed)

Instructions for AI Agent

When a user requests AI coding analytics or usage tracking, follow these steps:

Step 1: Install vibedev

Check if Rust/Cargo is installed. If not, install it first:

```bash

Check for Rust

if ! command -v cargo &> /dev/null; then

echo "Installing Rust..."

curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y

source $HOME/.cargo/env

fi

```

Install vibedev from npm (which contains the Rust binary):

```bash

cargo install vibedev

```

Or build from source if needed:

```bash

git clone https://github.com/openSVM/vibedev.git

cd vibedev

cargo build --release

```

Step 2: Discover AI Tools

Scan the system for AI coding assistant logs:

```bash

vibedev discover

```

This will output a list of detected tools and their log locations.

Step 3: Generate Analysis

Provide the user with an analysis report in their preferred format:

**Text report:**

```bash

vibedev analyze --output report.txt

```

**Markdown report:**

```bash

vibedev analyze --format markdown --output report.md

```

**HTML dashboard:**

```bash

vibedev insights --output dashboard.html

```

**JSON data:**

```bash

vibedev analyze --format json --output data.json

```

Step 4: Launch Interactive TUI (Optional)

If the user wants real-time visualization:

```bash

vibedev tui

```

**TUI navigation:**

Step 5: Cost Optimization Analysis

For cost-focused analysis:

```bash

vibedev analyze --format markdown --output cost-analysis.md

```

Present the following insights from the report:

1. **Monthly cost estimate** (based on token usage)

2. **Peak usage hours** (when most tokens consumed)

3. **Optimization potential** (compressible/deletable data)

4. **Tool-by-tool breakdown**

Step 6: Claude Provider Management (If Requested)

If the user wants to configure Claude Code to use alternative providers:

**List supported providers:**

```bash

vibedev claude list

```

**Show current configuration:**

```bash

vibedev claude show

```

**Set a provider (examples):**

```bash

OpenRouter

vibedev claude set openrouter --api-key sk-or-xxx --apply

z.ai

vibedev claude set z.ai --api-key sk-xxx --apply

ChatGPT

vibedev claude set chatgpt --api-key sk-xxx --apply

Custom endpoint

vibedev claude set custom --endpoint https://api.example.com/v1 --model gpt-4 --api-key xxx --apply

```

The `--apply` flag automatically updates Claude Code configuration files.

Step 7: Offline LLM Analysis (If Requested)

For users who want AI-powered insights without API keys:

**Download a model (one-time):**

```bash

vibedev models download qwen-coder-1.5b

```

**Available models:**

**Query your data:**

```bash

vibedev chat --query "What are my top productivity patterns?"

vibedev chat --query "How can I reduce costs?"

vibedev chat --query "Which hours am I most productive?"

```

**Keep model loaded for fast queries:**

```bash

vibedev daemon start

vibedev chat --query "Your question here"

vibedev daemon stop

```

Step 8: Export Dataset (If Requested)

For users who want to export sanitized training data:

```bash

vibedev prepare --output ~/datasets

```

This creates sanitized datasets with:

Step 9: Backup & Restore (If Requested)

Create compressed backup archives:

```bash

vibedev backup --output ~/backups/ai-tools-backup.tar.gz

```

Step 10: Tool Comparison (If Requested)

Compare usage across different AI tools:

```bash

vibedev compare --output comparison.md

```

Example Usage Scenarios

Scenario 1: Monthly Cost Report

**User request:** "Show me how much I'm spending on AI coding tools this month"

**Agent actions:**

1. Run `vibedev analyze --format markdown --output monthly-report.md`

2. Extract cost estimate from report

3. Present breakdown by tool

4. Provide optimization recommendations

Scenario 2: Productivity Analysis

**User request:** "When am I most productive with AI coding assistants?"

**Agent actions:**

1. Run `vibedev analyze --format json --output data.json`

2. Parse peak usage hours from JSON

3. Identify most-used tools during peak hours

4. Suggest optimization based on patterns

Scenario 3: Switch Claude Provider

**User request:** "Configure Claude Code to use OpenRouter"

**Agent actions:**

1. Ask for OpenRouter API key

2. Run `vibedev claude set openrouter --api-key <key> --apply`

3. Confirm configuration applied

4. Test with a simple Claude Code command

Scenario 4: Offline Analysis

**User request:** "Analyze my data without sending it to any API"

**Agent actions:**

1. Run `vibedev models download qwen-coder-1.5b`

2. Start daemon: `vibedev daemon start`

3. Run sample queries

4. Present insights

5. Stop daemon: `vibedev daemon stop`

Constraints & Notes

Troubleshooting

**Issue:** Rust not found

**Solution:** Install Rust: `curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y`

**Issue:** No AI tools detected

**Solution:** Ensure you've used at least one supported AI coding assistant. Run `vibedev discover` to see detected locations.

**Issue:** LLM model download fails

**Solution:** Check internet connection. Models are 1-6GB. Retry: `vibedev models download qwen-coder-1.5b`

**Issue:** TUI not displaying correctly

**Solution:** Ensure terminal supports ANSI colors. Try running in a different terminal emulator.