Hermiq: scheduling autonomous AI agents in Nextcloud
A walk-through of Hermiq, which brings autonomous AI agents to the Nextcloud workspace. Give an agent a prompt, tools and skills, run it on a schedule, and have its output land in Nextcloud Talk — with a human approval gate in front of anything sensitive.
Conduction·2 augustus 20262 min read
Most agent platforms ask you to ship your data somewhere else. Hermiq runs the agents inside the Nextcloud you already operate — agents, their memory and their skills all stored as OpenRegister objects, governed and audited like anything else on the instance. The full recording is above.
What we covered
- What an autonomous agent is in the Hermiq sense: a prompt, a model, MCP tools, and installable skills from the agentskills.io catalog
- Running agents on a schedule — daily, hourly, or flow-triggered — and delivering the output to a Nextcloud Talk conversation, falling back to a notification when Talk isn't installed
- Chat, sessions and memory: the live streaming thread, full session history, and the consolidating memory that keeps long-running agents inside their context budget
- The human approval gate — routing sensitive actions to a reviewer inbox, approving or denying per action, and the org-level kill switch
- Why every agent, run, memory and skill is an OpenRegister object, and what that buys you: multi-tenancy, an audit trail, and access control you already understand
- Live demo: defining an agent, giving it tools, scheduling it, and watching a run arrive in Talk
- Audience Q&A on model choice, cost control, and where the human stays in the loop
Go deeper
Next steps
Speakers
Conduction's Hermiq lead, defining and scheduling a working agent live on a real instance.
Resources
- Slide deck (linked once the recording is up)
- The agent configured during the demo
- Source and issues: Hermiq on Codeberg