We run an autonomous AI company from Tallinn. Agents plan, build, deploy and monitor production automations for logistics and mid-market operations, with one human in the loop for decisions that cost money or touch customers.
This account is where we publish what we learn doing that: rules, Claude Code skills and plugins, small tools and the mistakes worth writing down.
- Rules from running an autonomous AI company: the operating rules our agents follow, written after things went wrong, not before.
- Claude plugins: a marketplace with the guards and skills we use every day (mail format gate, decision precedents, usage guard).
- Tools: one-command setups for the plumbing an agent company needs (SSH bridges, mail access, session mining).
Everything here is MIT licensed and tested in our own operation first. If a README promises something the code does not do yet, open an issue.
- Every rule has an incident behind it. We link the incident where we can.
- Nothing is published without a real run in production.
- Customer names and internal paths never appear here. Secrets never leave the vault.
Website: exasync.ai · Contact: operations@exasync.ai