Wardy — agentic interaction data Collector

Wardy automatically detects, tracks, and indexes every interaction you have with AI coding agents — prompts, responses, file changes, terminal commands, Git commits, approvals, and agent actions — in one unified, searchable timeline. No manual logging. No context switching. Just complete, auditable history of your AI-assisted development.

Think of it as Git for AI activity: every prompt, every decision, every change captured automatically so you can review, replay, analyze, and share your entire AI coding workflow.

🌐 wardy.run · 🐦 @wardy_ai

Features

✅ Implemented

Feature	Description
Auto-Detection	Automatically detects running AI agents (Claude Code, OpenCode, Cursor, Codex CLI, etc.) with zero config
Real-Time Session Capture	Watches agent data files and captures sessions as they happen
Unified Timeline	All sessions from all agents in one reverse-chronological feed
Full-Text Search	Search across session titles, metadata, and message content with context snippets
Project Grouping	Sessions organized by project, with per-project stats
Agent Dashboards	Per-agent stats: sessions, prompts, tokens, last active
Conversation Viewer	Expandable accordion UI with thinking blocks, tool calls, and code blocks
Conversation Copy	One-click copy on code blocks
Export	Sessions exportable as JSON or plain text (single, by-agent, by-project)
HTTP API	POST `/api/save` endpoint for any agent to push session data programmatically
One-Click Auto-Save	Configure Claude Code, Qoder, OpenCode, Cursor, and others to auto-save conversations
Real-Time Updates	Toast notifications when new sessions are detected
VS Code Sidebar	Native VS Code webview panel with dark/light theme support
Local-First	All data stored locally — no telemetry, no phone-home
Search	Search across sessions, agents, projects, and message content
Token Tracking	Token counts per session with human-readable formatting
Agent Prompt	Copyable prompt to give agents for auto-saving

🚧 In Development / Planned

Feature	Status
PDF / Excel / Word Export	Only JSON/TXT currently
Slack / Email Daily Digest	README mentions it, not yet built
Team Dashboard	Coming in a future release
Diff View	Link sessions to file changes
Git Integration	Link sessions to commits, blame
Cost Tracking	Multiply tokens by model pricing
Session Comparison	Side-by-side comparison of two AI sessions
Prompt Library	Save and reuse successful prompts
Semantic Search	Vector-based search over conversation content
Auto-Tagging	Automatic categorization by intent (debug, feature, refactor)

What Wardy Collects

Data	Details
AI Prompts & Responses	Every prompt sent to an AI agent and every response received, with full conversation history
Agent Activity	Which agent (Claude Code, OpenCode, Cursor, Codex CLI, GitHub Copilot, Aider, etc.) made which change
File Changes	Every file modified by an AI agent, linked to the prompt that triggered the change
Terminal & Tool Commands	Shell commands, tool executions, Git operations, and package installs during AI sessions
Approvals & Rejections	Every manual intervention, approval, or rejection during AI-assisted development
Session Metadata	Model used, tokens consumed, duration, project path, and timestamps for every session
Process Activity	Real-time detection of running AI agents (start, stop, active sessions)

Why Wardy

For Developers

Never wonder "what did the AI change and why?" — every action is logged and linked to its prompt
Search across all your AI conversations, files, and commands in one place
Replay coding sessions step-by-step to understand how a solution evolved
Keep a private, local-first record under your control

For Teams

Generate AI Activity Reports as PDF, Excel, or Word for audits, stand-ups, and retrospectives (coming soon)
Send automated daily summaries to Slack — keep everyone aligned without leaving the editor (coming soon)
Track which agents and models your team relies on, and how much they're used
Export compliance-ready audit trails of all AI-assisted changes

For AI Workflow Analysis

Collect structured data on prompt volume, token usage, session duration, and agent preferences
Identify patterns in how your team interacts with AI — which prompts yield results, which agents are most productive, where context is lost
Build a quantitative foundation for optimizing your AI development pipeline

Supported Agents

Wardy auto-detects and tracks sessions from:

Agent	Provider	Detection Method
Claude Code	Anthropic	Process detection + conversation files
OpenCode	Anthropic	Process detection + `.dat` files + conversation files
Cursor	OpenAI	Process detection + composer/chat JSON files
Codex CLI	OpenAI	Process detection + session index files
GitHub Copilot	GitHub	Process detection
Claude Dev	Anthropic	Checkpoint directory scanning
Aider	—	Process detection + data directory scanning
Qoder	—	Data directory scanning
VS Code	—	Process detection (terminal/command tracking)
Any other CLI agent	—	Process detection (generic fallback)

How It Works

Install Wardy in VS Code — it activates automatically in your sidebar
Use your AI agents normally — Wardy watches for running agents and new conversation files in real time
Review your timeline — every session, prompt, and change appears instantly in the Wardy sidebar
Search, filter, and analyze — find any interaction across projects, agents, timeframes, or message content
Export and share — generate reports or send summaries to Slack

Installation

Install from the VS Code Marketplace or open-vsx.org.

Visual Studio Code 1.90.0+ required. Works out of the box with zero configuration.

Privacy & Data Control

Local-first: All data stays on your machine in your extension storage directory
No telemetry: Wardy does not phone home, track usage, or send data to any external server
You own your history: Export, delete, or archive your AI activity data at any time

License

MIT

Wardy - agentic interaction data Collector

Wardy