Three small tools for your workstation: your AI history, a context meter, a fleet inventory

Not every tool needs a server. These three are single scripts, Python standard library only, that each answer one everyday question: where are my old conversations, how close is my agent to forgetting, and what moved on my machines.

What you get. Your whole AI chat history as files you can read, search and index; a context gauge for a local agent, which no hosted plan shows you; and a weekly "what changed" for every machine you own.

1. Your Claude.ai history, as files: claude-archive

The official export (Settings → Privacy → Export data) is complete but raw JSON. One script turns it into an archive you can use:

python3 claude_archive.py ~/Downloads/data-XXXX.zip -o ./claude-archive

• one Markdown file per conversation, in date order;
• every code block extracted as a real source file;
• a checklist of attachments to fetch by hand, since the export does not contain them;
• a manifest with input and output counts: nothing is dropped silently.

No network, no dependency. Ideal to feed your own search over your notes.

2. A context meter in the status bar: openfox-omarchy

A hosted agent shows what is left of your plan. A local one has no plan; its real limit is the context, which fills up and drops at every compaction. A small script reads OpenFox's session store and writes the record that Omarchy's agents widget already draws, next to Claude Code and Codex:

install -m 755 openfox-usage ~/.local/bin/openfox-usage
openfox-usage      # then a user timer every ten minutes

When you add a provider to the widget's settings, list all of them: giving providers replaces the default list, so leaving out claude hides Claude.

3. Inventory by scanning, not declaring: parc

Ansible pushes a state and forgets it; monitoring tools observe without comparing. parc starts from what exists: you scan your machines, and it tells you what moved.

./parc.py scan local laptop server   # snapshot of each machine
./parc.py diff                       # what changed since
./parc.py compare laptop server      # gaps between two machines
./parc.py since 7d                   # what moved this week
./parc.py export-ansible             # inventory and a reconciliation playbook

Read-only by default; apply only installs after a confirmation typed by hand, and never uninstalls. diff exits with 1 when a machine strays, so it fits a cron.

Found this useful? Tell us what to measure next.