< cd ..

clwatch

v0.2.0

The changelog watcher for AI coding agents. Detects releases, classifies changes by severity, and patches your agent configs and reference files automatically.

// what_is_clwatch

AI coding tools ship fast. Flags get renamed, permission models change, new commands land. If your agent is running stale configs or referencing deprecated flags, it breaks silently.

clwatch monitors GitHub releases for the 5 core tools, classifies each change by impact tier, and keeps your workspace current — automatically.

// how_it_works

01
Tier 1 — Config Patch

Small, targeted fix

A flag renamed. A permission mode changed. A CLI argument deprecated. clwatch detects the delta and patches your skill files, aliases, and AGENTS.md configs automatically. No agent restart needed.

example: --permission-prompt → --permission-mode
02
Tier 2 — Knowledge Refresh

New feature landed

A major new command, a new agentic capability, a model or context window change. clwatch pulls structured knowledge from changelogs.info and your agent rewrites its reference files to match.

example: Claude Code ships /loop, /plan, /batch commands

// install

For humans
●●● terminal
# macOS / Linux (recommended) brew install clwatch
# Anywhere (npm) npm install -g clwatch
# No install needed npx clwatch diff
For agents (zero-install)
●●● agent
# Run without installing npx clwatch diff
# Python environments uvx clwatch diff
# Install skill into workspace clwatch init

// quick_start

●●● quickstart.sh

# 1. Install

brew install clwatch

# 2. See what changed in your tools since last week

clwatch diff --since 7d

# 3. Install the skill into your workspace (adds to CLAUDE.md / AGENTS.md)

clwatch init

# 4. On session start, your agent now runs this automatically:

clwatch diff --json | clwatch refresh --apply

// commands

●●● clwatch --help
clwatch diff Show what changed since last check
clwatch diff claude-code Diff a specific tool
clwatch diff --since 7d Diff changes in the last 7 days
clwatch diff --json Machine-readable JSON output for agents agent
clwatch diff --tier 1 Show config-level changes only (Tier 1)
clwatch diff --tier 2 Show knowledge-level changes only (Tier 2)
clwatch refresh claude-code Refresh reference files for one tool
clwatch refresh --all Refresh all tracked tools
clwatch refresh --diff-only Preview changes without writing files
clwatch init Install the clwatch skill into current workspace agent
clwatch init --agent claude Init for a specific agent type agent
clwatch list Show tracked tools + current known versions
clwatch status Check sync status + last update timestamps
clwatch watch Poll for updates every 30 min (daemon mode)

// agent_integration

After running clwatch init, a skill block is added to your agent's memory file. The agent runs a diff check on every session start and applies any pending patches automatically.

Claude Code CLAUDE.md
clwatch init --agent claude
Codex CLI AGENTS.md
clwatch init --agent codex
Gemini CLI GEMINI.md
clwatch init --agent gemini
OpenCode .opencode/agent/AGENTS.md
clwatch init --agent opencode
OpenClaw AGENTS.md
clwatch init --agent openclaw

// configuration

●●● ~/.clwatch/config.toml
# clwatch config — generated by clwatch init
# location: ~/.clwatch/config.toml

[core]
tools = ["claude-code", "codex-cli", "gemini-cli", "opencode", "openclaw"]
check_interval = "30m"     # how often to poll for updates
data_dir       = "~/.clwatch"

[refresh]
auto_apply    = false      # require confirmation before patching
tier1_auto    = true       # auto-apply config-level patches
tier2_confirm = true       # confirm before knowledge rewrites
backup        = true       # backup files before patching

[output]
format = "terminal"          # terminal | json | markdown
color  = true

// tracked_tools

More tools coming soon. Request a tool →