1
Human decision-maker
Names the outcome, approves anything public/paid/irreversible, and keeps judgement where it belongs.
Get started / DIY
Want to copy this? Start here.
The path is simple: set up Multica, copy the starter files, create the four-agent pod, prove one useful workflow, then add Hermes as the persistent operator layer.
The DIY path
Start with the board. Copy the files. Then add the operator.
This page should not make visitors reverse-engineer the sequence. Multica comes first because it gives the work a visible board. The copyable files give the pod usable roles and behaviours. Hermes comes after that when the loop needs memory, tools, schedules, and channels.
- 1
Set up Multica
Start with the visible work board: install/authenticate Multica, connect the local daemon/runtime, and confirm Claude Code can execute agent work from the same machine.
- 2
Copy the starter files
Use our role files, skill files, setup instructions, workflow recipe, briefs, review checklists, and approval gates instead of writing the whole operating layer from scratch.
- 3
Create the four-agent pod
Create Chief of Staff, Researcher, SEO Consultant, and Blog Writer. Paste the role files, attach the core skills, and keep the pod small enough to understand.
- 4
Run one useful request
Use the SEO/site-growth demo to prove the loop: one human request becomes a parent issue, specialist tickets, handbacks, readable reports, and a final recommendation.
- 5
Add Hermes as operator
After the Multica loop works, add Hermes for persistent memory, tools, messaging, schedules, MCP, and coordination that survives one terminal session.
What you are building
What each layer is doing.
The setup is not “install every shiny thing.” It is a layered system: a human decision-maker, a coordinator, specialists, a visible board, and then tool-connected runtime when the basic loop works.
2
Operator / Chief of Staff
Turns intent into clear work: owner, success signal, inputs, blockers, approval gate, and final handback.
3
Specialists
Do narrow work in named lanes: research, SEO, writing, creative, build, analysis, support, partnerships.
4
Workflow board
Keeps tasks visible. In this starter path, Multica is the board: issues, assignees, comments, status, artifacts.
5
Tool-connected runtime
Makes agents actually useful: files, browser, terminal, GitHub, Slack, email, calendars, cron, MCP, and internal systems.
Build order
Multica first. Hermes once the loop is real.
1. Multica + starter files
- Install Claude Code and Multica on one Mac/PC.
- Use Multica as the visible issue board.
- Copy the YDY role files and skills from Playbooks.
- Create the four-agent starter pod.
- Run one request through tickets, specialists, reports, and final review.
2. Hermes operator
- Add persistent memory and reusable skills.
- Connect channels like Telegram, Slack, email, or Discord.
- Connect tools, cron, MCP, browser, files, GitHub, and internal systems.
- Let Hermes coordinate above the Multica specialists when the workflow needs to persist.
Copyable pack
Use the Playbooks page when you are ready to copy files.
The Playbooks page is the copyable asset pack: agent roles, skill files, setup instructions, workflow recipes, briefs, checklists, and approval gates. Use it alongside the Multica guide.