What do you want to set up?
Set up website logins and API keys
Choose the right store, save one low-risk credential, test the connection, and know exactly how to revoke it.
You’ll finish with: A working website login or centrally managed API key, and a tested way to revoke it.
Connect one GitHub repository with a read-only token
Create a narrowly scoped credential, prove that a second private repository is inaccessible, and revoke it after the job.
You’ll finish with: A verified one-repository connection and a repeatable revocation check.
Open your remote WebUI through an SSH tunnel
Reach an existing Hermes WebUI from your laptop while port 8787 stays bound to the server’s loopback interface.
You’ll finish with: A working local browser URL, plus a check that closing the tunnel stops that route.
Back up Hermes and rehearse a restore
Make an encrypted copy of a stopped native install’s state, restore it somewhere empty, and compare it before updating.
You’ll finish with: An encrypted archive and a verified separate restore directory.
Give an agent read-only input and a writable output folder
Use a disposable profile and Docker terminal backend to test a narrow file boundary before exposing a real workspace.
You’ll finish with: A test that writes the report, fails to write to the input folder, and leaves the input file untouched.
Trace where one agent message goes
Use a harmless marker to inspect the model, tool and memory routes before sending a real private document.
You’ll finish with: A filled-in data-flow checklist for your actual profile, with a test that reveals automatic memory sync.
Measure and cap the cost of a scheduled agent job
Run one representative task, calculate a monthly budget, and bound retries before you put it on a schedule.
You’ll finish with: A cost-per-run estimate, a schedule budget, and explicit iteration and retry limits.
Start with a working installation
These recipes assume you can already run Hermes. Follow Setup, choose a model, and send a first test message. Developers can connect through the Bridge API; the memory guide explains the available memory routes.
Use synthetic data and a test profile for the first run.