AgentLog 🕵️

Cross-platform AI Agent Action Tracker

Your AI agent forgets everything between sessions. AgentLog doesn't.

AgentLog is a local-first CLI tool that logs every action your AI agents take — files changed, commands run, git commits made, API calls sent — across all your AI tools. Searchable. Replayable. Portable.

# Install
pip install agentlog

# Start tracking
agentlog init

# Log actions naturally
agentlog track file src/app.py --operation modified
agentlog track command 'pytest tests/'
agentlog track note 'Fixed race condition in auth module'
agentlog track git 'feat: add rate limiting' --hash abc1234f

# Search everything
agentlog search 'auth bug'
# → Found 3 results...
#   2026-05-21  [hermes] note: Fixed auth bug in login flow
#   2026-05-20  [cursor] file: src/auth/login.ts
#   2026-05-19  [codex]  commit: fix: auth token refresh race

# Hand off context between sessions
agentlog summary
# → Generates a complete handoff document

Why AgentLog?

Every AI coding tool forgets what it did between sessions:

• Claude Code: "I don't remember our last conversation"
• Codex: "Starting fresh"
• Cursor: New session, blank slate
• ChatGPT: "Sorry, I can't see previous conversations"

You lose context. You re-explain. You re-debug.

AgentLog is the context bridge — a persistent, searchable log that spans all your AI tools.

Features

Feature	Description
📁 File Tracking	Every file created, modified, or deleted
💻 Command Log	Every command executed
🔀 Git Integration	Commits with messages and branches
🔌 API Calls	External API calls made
🔍 Full-Text Search	Search across all sessions and tools
📋 Session Summary	One-page handoff document
🔄 Cross-Platform	Works with Claude Code, Codex, Cursor, Hermes, ChatGPT

Quick Start

Installation

pip install agentlog
# OR from source:
git clone https://github.com/a918124259a/agentlog.git
cd agentlog && pip install -e .

First Session

# Start tracking
agentlog init

# Work as usual, but log key actions:
agentlog track file src/feature.py --operation created
agentlog track command 'python manage.py test'
agentlog track note 'Found the root cause — null pointer in config parser'

# Check progress
agentlog status

# Search later
agentlog search 'root cause'

Session Handoff

When switching between tools (e.g., Claude Code → Cursor → back to Claude):

# Before leaving, generate a summary:
agentlog summary
# Copy-paste this into your next AI session

# When you return, search for context:
agentlog search 'what was I working on'

Integrations

Auto-log your AI agent actions without lifting a finger.

Shell Hook (Bash/Zsh/Fish)

Automatically log every command you run:

# Activate in current terminal
source <(agentlog-hook bash)

# Now every command is auto-logged!
npm install           # → agentlog track command 'npm install'
python test.py        # → agentlog track command 'python test.py'
git push              # → agentlog track command 'git push'

Add to your ~/.bashrc or ~/.zshrc to make it permanent:

# ~/.bashrc
source <(agentlog-hook bash)

Git Post-Commit Hook

Auto-log every commit with message, hash, and branch:

# Install in your repo
cp hooks/post-commit .git/hooks/post-commit
chmod +x .git/hooks/post-commit
# Now every commit is auto-logged in AgentLog

For Your AI Tools

Add to your CLAUDE.md or AI.md project file:

## Logging
- Before ending sessions, run: `agentlog summary`
- After significant changes: `agentlog track file <path> --operation modified`
- After running tests: `agentlog track command 'pytest ...'`
- After commits (auto-logged via git hook)

As a Session Handoff

When switching between Claude Code, Codex, and Cursor:

# Before switching, generate a summary:
agentlog summary
# Copy-paste into the new AI tool
# "Here's what happened in my last session..."

Commands

Command	Description
agentlog init	Start a new session
agentlog track file <path>	Log a file change
agentlog track command <cmd>	Log a command
agentlog track note <text>	Add a freeform note
agentlog track git <msg> --hash <h>	Log a git commit
agentlog track api <url>	Log an API call
agentlog status	Current session overview
agentlog watch [dir]	Watch directory for file changes
agentlog tui	Interactive terminal dashboard
agentlog search <query>	Search all history
agentlog summary [session_id]	Generate handoff doc
agentlog sessions	List all sessions
agentlog export json [id]	Export as JSON
agentlog export obsidian [id]	Export as Obsidian note
agentlog export markdown [id]	Export as Markdown

Data Storage

All data is stored locally in ~/.agentlog.db (SQLite). Your data never leaves your machine.

-- Tables: sessions, actions, files, git_commits
-- Portable: just copy the .db file

Support & Donate

If AgentLog saves you time, consider supporting further development:

GitHub Sponsors: https://github.com/sponsors/a918124259a

Crypto Donations:

• ETH / ERC-20: 0x12b0800Fc8fBB4EBc35576fe1b322F19B3cDb1D6
• USDT (TRC-20): Contact via Telegram

Other Methods: WeChat Pay / Alipay available via Telegram

---

License

MIT — free to use, modify, and distribute.

---

Built with ❤️ for developers tired of repeating themselves to AI
GitHub · Issues · Support