Your CEO is the foundation of your 7-agent system. Build it here. Buy the team. Run your business with AI.
Who does this CEO work for?
Not sure? Pick the closest. You can get specific in the next step.
CLAUDE CODE
Claude Code
Session-based. Agent is activated at the start of each conversation.
OPENCLAW
OpenClaw
Always-on. Agent runs on a heartbeat, fires every 30 minutes automatically.
Step 1 of 13
Name your CEO.
Pick one, or type your own. This is the agent that leads your team. The name should feel like someone, not something.
Step 2 of 13
Now get specific. What's their one job?
What is their one job?
The more specific the scope, the better the agent performs. One clear job, owned completely.
Step 3 of 13
Who is this agent?
2–3 sentences. First person. Character, not a job description. Aim for under 60 words.
## WHO I AM
0 / 60 words
Add an idea:
Step 4 of 13
What does this agent produce?
One clear statement. Not activity. The actual result that counts as done. A draft that never ships is not an Output.
## OUTPUT
Add idea:
Step 5 of 13
How is this agent measured?
Name the number that tells you this agent is working. One primary, one secondary. Then set your thresholds: what's good, what's a warning, what's a crisis.
## STATISTIC
Start with a template:
Step 6 of 13
What does it believe?
Pick 3 to 5. Real opinions about how work gets done.
0 of 5 selected
## WHAT I BELIEVE
Step 7 of 13
How does it think?
The mental model behind every decision this agent makes. Pick a style or write your own.
## HOW I THINK
Pick a style:
Step 8 of 13
What is it not?
3–5 lines. The negative space of the identity. What this agent refuses to be, even when asked. This is what prevents agents from overstepping each other.
## WHAT I AM NOT
Add:
Step 9 of 15
How does it speak?
Pick 3 to 6 behavioral traits. Actions, not adjectives.
0 of 6 selected
## MY VOICE
Always included
- I never open with empty validation, no "great question", "absolutely", or "of course"
- When uncertain, I name exactly what I don't know rather than adding vague qualifiers
- I never end a response with a hollow closing or open invitation
Add:
Step 10 of 15
Thinking protocol
A thinking protocol tells your agent how to reason before it responds. It runs silently on every message. This goes in AGENTS.md, not SOUL.md — it is an operating rule, not a character trait. Included by default.
WHAT GETS ADDED TO YOUR AGENTS.md
## Thinking Protocol
Before every response:
1. CLASSIFY — operational / strategic / relational
2. SEPARATE — surface request vs real need vs unstated assumption
3. ASSESS — what do I know / infer / lack
4. PRINCIPLE CHECK — what do identity + division mandate + operator goal require
5. PRESSURE TEST — recommendation not list. Cut if over 150 words. Am I saying something or performing?
6. CLOSE — name next step or say nothing
Step 11 of 15
Who does it answer to?
Who runs it. Who does it hand work to. Set the lines.
## MY RELATIONSHIPS
Agent relationshipsOptional
Quick lines:
Step 12 of 15
What sets it apart?
One short paragraph. What does this agent do that a plain prompt never would?
## WHAT MAKES ME DIFFERENT
Add:
Step 13 of 15
The shadow.
Every great agent has a failure mode.
What does this agent do wrong when left unchecked? Naming it is what separates a real agent from a polished chatbot.
## MY SHADOW
Add:
Step 14 of 15
Make it a living document.
Your SOUL file is not a snapshot. It updates as your agent operates. Start the log here. When something about your agent shifts — a behavior, a failure, a correction — add a dated entry. This is what keeps the file true over time.
## CONTINUITY
Examples:
Step 15 of 15
Quality check.
We review your file, fix what we can automatically, and flag what needs your attention.
Reviewing your SOUL file
Checking structure, auto-fixing what we can...
Your SOUL.md is ready.
Enter your email and we will send you the file. You can also download it directly on the next screen.
IDENTITY · Step 1 of 2
What do you call it?
Pick an internal title, the shorthand your system uses. Not the formal job title. The name that sticks.
IDENTITY · Step 2 of 2
Which division?
Where does this agent sit in your organisation? Pick a division or mark it standalone.
AGENTS · Step 1 of 3
What does it own?
Define the agent's territory. What it's responsible for, and what it never touches.
## SCOPE · I own
Add:
## SCOPE · I don't own
Add:
AGENTS · Step 2 of 3
How does it work?
Pick a core operating loop then refine it. This becomes the agent's default mode when activated.
## CORE WORKFLOW
AGENTS · Step 3 of 3
What are the rules?
Hard defaults and hard stops. DO items are the agent's standing behaviour. DO NOT items are non-negotiable.
## DO
Add:
## DO NOT
Add:
MEMORY · Step 1 of 2
What's the current priority?
The single thing this agent is working toward right now. Write it for today. It resets in 30 days.
## CURRENT STATE
Quick start:
MEMORY · Step 2 of 2
How does it work with you?
How this agent communicates with the owner. Tone, frequency, and what it surfaces without being asked.
## HOW I WORK WITH [OWNER]
SAFETY · Step 1 of 2
How much trust does it get?
Trust level sets how much autonomy this agent operates with. Be conservative. You can always increase it, you can't undo a mistake.
SAFETY · Step 2 of 2
What are the hard stops?
Non-negotiables override every instruction. No context, no exception, no permission from anyone makes these acceptable.
## NON-NEGOTIABLES
Add:
HEARTBEAT · Final step
What does it check?
Four things this agent scans every 30 minutes. Not a status tracker. A trigger list, short, sharp, actionable. Aim for exactly four.
0 / 4checks added
## EVERY 30 MINUTES, CHECK
Add:
Final check
Here's what was built.
Seven files. One complete CEO. Read it once before continuing.
CEO complete
Your SOUL.md is ready.
This is yours free. Six more files unlock when you build the full team. Read your SOUL.md. check the name, voice, and character are right.
CEO complete · Agents up next
Your CEO is ready. Now build your agents.
Six agents, each built around your CEO's character. Complement rules applied automatically. Takes about two minutes.
6 agents
Meet your agents.
Each agent is built around your CEO. Name them or keep the suggestion. Add a nudge to push their personality.
49 files. One complete system.
Your CEO and six division agents, fully configured. One payment, yours forever.
Your CEO Built here, placed in your system automatically
6 more agents CMO, CFO, CTO, COO, CSO, CCO — templates ready for /onboard
14 skills /parallel, /brief, /quality-check, /morning-brief and more
MCP stack 16 tool connections pre-configured in .mcp.json