Case study / Personal system · Local Docker
Tim AI Hermes.
A personal AI workflow I can inspect.
Built around Hermes to support invoices and project progress reports, with Mission Control providing monitoring and a visual view of the team.
- My role
- Idea through local deployment
- Technology
- Hermes · Next.js · Three.js · SQLite · Docker
- Project context
- Personal system · Local Docker
01 / Tim AI Hermes
The starting point
I built a personal AI team around Hermes to support my work, including client invoices and progress reports for ongoing projects.
The workflow involves roles, task states, project context, and work that may need attention. I wanted a shared view of those signals so I could inspect the team’s current state and follow the context behind a task.
Hermes provides the agent framework. My project brings together the personal team setup, its working process, and Mission Control as a local monitoring interface.
02 / Tim AI Hermes
My contribution
Idea & planning
Defined the purpose of the personal AI team and planned how the workflow and monitoring interface should fit together.
Agent workflow
Configured the team setup, agent workflow, and supporting environment around Hermes.
Mission Control development
Built the dashboard, data views, and virtual-office interface for inspecting the team and recorded work.
Testing & local deployment
Tested the system and prepared its Docker deployment on my own machine.
This is an integration and workflow project built around Hermes. The current deployment is local through Docker; my confirmed practical uses are invoices and project progress reports.
03 / The workflow
Follow the work and its context.
- 01
Set up the team
Define roles and configure the underlying agent workflow.
- 02
Use the workflow
Prepare invoices and project progress reports.
- 03
Inspect the state
Review agent activity, task ownership, and work that needs attention.
- 04
Review the context
Open recorded task descriptions and event timelines in Mission Control.
04 / Tim AI Hermes
A monitoring interface with clear boundaries
Give monitoring its own boundary
Mission Control is read-only. It displays tasks and runtime information while operational changes remain in the underlying workflow. A separate service, restricted data mounts, information filtering, and session-based access support that boundary.
Distinguish roles from current activity
The roster shows an agent’s role and configured status alongside its live state. Task views add ownership, descriptions, status, and recorded events. Idle and Blocked remain visible because they are useful supervision signals.
Keep project context connected
Organisation and project filters provide context across views. Task details and event timelines help me inspect how work progressed beyond a final status label.
Make the same status spatially readable
The virtual office presents rooms, desks, avatars, and status indicators, with a 2D floor-plan alternative. Avatar activities are a visual layer: coffee, rest, or prayer animations do not pause an agent process or change its task state.
05 / Inside the application
Team status, task history, and a virtual office.
Agent cards connect roles, runtime states, and task context. The view includes idle or blocked work when that is what the system records.
The selected task is a personal Tando design task recorded as Done. Its description and timeline provide context for reviewing the work.
The same team can be explored through a 2D floor plan alongside the 3D office shown above.
Real local Mission Control data, captured with the owner’s permission on 20 October 2026. Statuses may change. Selected views show agent roles and a personal Tando task; they do not include client financial documents, credentials, or raw agent conversations.
06 / Tim AI Hermes
Outcome & scope
The project gives my personal Hermes workflow an interface for monitoring and review, alongside its practical use for invoices and progress reports. It also supports ongoing development and testing of additional agent roles and workflows.
The system currently runs locally through Docker. Screenshots are operational snapshots, including onboarding work and tasks needing attention; they are not evidence of measured time savings or fully autonomous project delivery.
A conversation starts here
Have an idea or a collaboration in mind?
Let’s talk it through.