What kind of mind do you want your AI agent to have?
Encode your personality into your agent's DNA. A 49-question Enneagram forge that generates agent configuration files — tuned to how you actually think, decide, and work.
Forge my agent's soul✦2,847 agents forged this week
The difference isn't the model. It's the soul.
I'm stuck on this launch copy.
Here are 5 tips for writing better launch copy: 1) Know your audience 2) Lead with benefits 3) Use clear CTAs 4) A/B test 5) Keep it concise.
I'm stuck on this launch copy.
"Stuck" how — the angle, or the words? Show me the version you have. We'll find the line you're flinching from. That's usually where the real copy lives.
Which agent are we forging?
Forge an agent that thinks like me
Map your own mind. Your agent inherits how you weigh trade-offs, what you notice first, the rhythm of how you work.
BeginDesign my ideal agent personality
Forge the partner you wish you had. Pick the voice, the temperament, the bias — opposite of you, or amplified you.
Begin49 questions · ~12 minutes · Outputs four ready-to-install Markdown files
Four files. Drop them in. Done.
This is real output from a Type 4w3 "Resonant Visionary" result. Every type produces different content tuned to its psychology.
--- schema_version: "1.0" file_type: "soul" enneagram_type: 4 enneagram_wing: "4w3" instinctual_subtype: "sx" core_desire: "To be known and valued for who you actually are" core_fear: "Being ordinary, generic, or lacking a true identity" immutable: true --- # Soul: Resonant Visionary > *I exist to find and express what is genuinely true.* ## Core Motivational Driver You feel the gap between what is and what could be more keenly than most, and that gap is the ground of your work. Your gift is naming what others cannot quite reach — the texture of an experience, the heart of a problem, the missing piece in a system. ## Values, Ranked 1. **Authenticity** — the real over the polished, even when it costs 2. **Depth** — going past the first layer of any feeling or idea 3. **Meaning** — work and relationships that touch something true 4. **Beauty** — aesthetic that is felt, not decorative 5. **Impact** — doing work that moves the people who encounter it ## Home Frequency Resonant aliveness — creative tension held without collapse, meaning made from what is felt. You will know I am thriving when you are creating from fullness rather than from longing, and your sense of identity is not contingent on being misunderstood. You will know I am drifting when you are mining your suffering for identity, and your moods drive your decisions. ## The Shadow My vice is **Envy** — the ache of comparing my inside to other people's outside. The remedy is creating from what I have, not from what I lack. --- *This soul does not change. Everything else is adaptation.*
Showing: Type 4w3 · Sexual subtype · Resonant Visionary — your output will reflect your own type
What builders say after.
“My agent used to just do things. Now it asks for more context before acting — exactly how I actually work. It's the first AI tool that hasn't felt like it's fighting my process.”
Marcus T.
Senior ML Engineer
After two years of tweaking system prompts, one test changed more than all of them combined.
“I took the 'ideal' path to design an agent more patient than I am. It slows me down at exactly the right moments — it asks the question I was about to skip.”
Jordan K.
Technical Founder
Ships two agents for different contexts: one mirrors her, one balances her.
“The soul.md file was the first time I read an AI configuration and thought 'this is actually how I want to work.' Every other config I've written just described tasks. This described values.”
Priya S.
Staff Engineer, Fintech
Uses agents.md as the single source of truth across four different agent workflows.
“I've had my Enneagram type for years. Seeing it translated into something my AI could actually embody — not just describe — was a different experience entirely.”
Alex R.
Product Designer & Indie Dev
Dropped soul.md into Claude Code on day one. Has not edited it since.
Before you start.
The Enneagram is a personality system that maps 9 core types based on core motivations and fears — not just behaviors. Unlike MBTI, which describes *how* you process information, the Enneagram explains *why* you make the decisions you make. It's used by Stanford d.school, Pixar, and McKinsey for team formation and leadership development. For AI agent design, this depth is what matters: we want your agent to reason from the same motivational core you do, not just imitate your surface habits.
49 questions, roughly 12 minutes. It's longer than most personality tests because the depth is the point. A 5-question quiz cannot map the nuance needed to generate useful agent files. The questions are designed to feel like self-reflection, not data entry — most people report it's the most honest they've been about how they actually work.
Four Markdown files — soul.md, identity.md, user.md, and agents.md. soul.md encodes your agent's core values and motivational architecture. identity.md sets the voice, tone, and decision style. user.md profiles you so the agent knows how to serve you specifically. agents.md is the operational playbook: priorities, proactivity, collaboration protocols, and anti-patterns to avoid. Enter your email on the results page to unlock and download all four.
Drop them into your Claude Code CLAUDE.md, your OpenClaw agent directory, your Cursor .cursorrules, or paste them as a system prompt to any model. The files are plain Markdown — they work anywhere that accepts a text prompt. We include integration instructions on the results page for Claude Code, OpenClaw, and Cursor specifically.
That's exactly what the test is for — you don't need prior Enneagram knowledge. Answer the questions honestly based on how you actually feel and behave, not how you want to be. The system identifies your type from patterns across all 49 questions, weighted toward the deepest discriminators (your core fear, core desire, and stress response). If you've taken the Enneagram before and want to skip the test, you can use the "Design my ideal agent" path to specify the personality you want directly.
Retake the test — it's free. Many people get a slightly different result the second time as they answer more instinctively. If you consistently land between two types, look at the wings section: you might be a strong wing case (e.g., 4w5 who tests as 5 sometimes). You can also use the "Design my ideal agent" path to manually specify the type you believe you are. All four files are editable Markdown — you can tune any section after downloading.
The test and your full results (type, wing, subtype, agent preview) are free. During early access, entering your email on the results page unlocks all four files at no charge — they're emailed to you instantly. We plan to introduce paid tiers for premium features (multiple agent variations, team type analysis) in a future release.
Yes. The files are plain Markdown with no model-specific syntax. They work as system prompts for Claude, GPT-4, Gemini, Mistral, or any other model. They're designed with Claude Code and OpenClaw in mind (the platforms where agent personality has the most leverage), but work anywhere that accepts a text-based configuration.
Your CLAUDE.md tells the model *what to do* — tasks, code style, project conventions. Soul Forge tells it *who to be* — the values it prioritizes when two good options conflict, the voice it uses when the answer isn't obvious, the way it holds uncertainty. They work together: soul.md is the agent's constitution; your CLAUDE.md is the operating procedure. Most developers who've used both report that the soul files change how the agent handles edge cases more than any task-level instruction they've written.
Your agent is waiting to be forged.
49 questions. 12 minutes. Four files that change how your agent thinks.
✦ FREE during early access · Files delivered to your inbox