Make AI work like you think.
Mirror Agent is a configuration pack that turns ChatGPT or Claude into a consistent operator that follows your rules, speaks your language, and refuses what you'd refuse.
This is for you if:
- You use AI daily but hate re-explaining your preferences
- You want consistent tone and behavior across sessions
- You need guardrails—things the AI should never do
- You've tried custom instructions and hit their limits
- You want presets for different modes (build, market, roast)
This is NOT for you if:
- You've never used ChatGPT or Claude
- You want a hosted AI product, not configuration files
- You expect AI to read your mind without setup
What AI feels like without a system
Every conversation starts from scratch. Here's the cost:
Re-explaining your preferences every single session
Inconsistent tone—helpful one day, annoying the next
AI ignores your constraints when context gets long
No way to enforce 'never do X' rules reliably
Switching between tasks means re-prompting everything
Outputs that sound nothing like you
How Mirror Agent works
A modular config system that shapes AI behavior
Load the constitution
Non-negotiable rules: identity, priorities, what to refuse, output standards. These never change.
Select a preset
Discovery mode, build mode, marketing mode, roast mode. Each preset loads different tone and task fragments.
Customize switches
Roast on/off, spice level, citation strictness, language defaults. Toggle without rewriting prompts.
Run with consistency
Same behavior across ChatGPT, Claude, or any capable LLM. Your operator, your rules.
What's inside the pack
Everything you need to configure a consistent AI operator
System Constitution
Core identity, priorities, evidence rules, output standards, refusal policies.
Persona Profile
Your operator's traits, language rules, memory policy, and mirror principles.
Prompt Fragments
Reusable blocks: tone packs, task prompts, mode configs. Mix and match.
Presets Library
Ready-to-use profiles: discovery, build, marketing, default. One-click mode switching.
Config Switches
YAML toggles for roast mode, spice level, citations, language. No prompt rewriting.
Response Templates
Output shapes: structured answers, short answers, roast mode, evidence packs.
Bootstrap Prompt
Copy-paste to start any chat with your full config loaded.
Runtime Adapters
Loaders for OpenAI API, Claude API, Cloudflare Workers.
What's NOT included:
Config over chaos
Modular. Consistent. Yours.
4
Preset modes
6
Constitution rules
~5 min
Setup time
Any
LLM compatible
[Discovery] Running validation cycle...
[Score] Candidate passed: 18/24
[Result] Moving to experiment phase
Ready to build.
Simple pricing
One pack. One price. Yours forever.
Mirror Agent Pack
Everything you need to configure AI that works like you think.
- Full system constitution
- Persona profile templates
- 4 preset modes (discovery, build, marketing, default)
- Prompt fragment library
- Config switches (YAML)
- Response templates
- Runtime adapters
- Lifetime access + updates
Questions
No. It's configuration files you load into ChatGPT, Claude, or any LLM. You bring your own AI access.
Yes. The system prompt and constitution can be pasted directly into custom GPT instructions, Claude projects, or any system that accepts custom prompts.
That's the point. The pack gives you the structure—constitution, fragments, switches. You edit them to match your thinking style.
Not for basic use. Copy-paste the bootstrap prompt and go. The runtime adapters need basic familiarity with APIs if you want programmatic access.
Custom instructions are one blob of text. Mirror Agent is modular: separate constitution, persona, tone packs, task fragments, and switches. You can change modes without rewriting everything.
Stop re-explaining yourself.
Get the pack. Load it once. AI that works like you think.
Get the Pack — $10One-time purchase. Instant download. No subscription.