1.
Message your agent.
On Telegram, WhatsApp, Signal, Discord, or wherever you already talk to your OpenClaw or Hermes agent, just say mirror session on [what's on your mind].
Chat with yourself.
Mirror is an AI agent that simulates being you so you can have a deeper conversation.
How it actually works
2.
Pick a mode.
The agent asks how you want this one to go: brainstorm, analyze, decide, process, future self, or inner critic. Each one is a different voice from inside your own head.
3.
Talk to yourself.
The agent pulls context from your gbrain so the other-you sounds like you, then plays that side of the conversation in first person. Text or voice, whatever channel you're on.
4.
It saves to your brain.
When you wrap, the full transcript goes into gbrain as a mirror-session page. Your inner dialogue becomes part of who your agent remembers you to be.
Six versions of you.
Pick a voice for the other side of the conversation. They each listen differently.
Brainstorm
generate ideas with yourself
Build on ideas, connect things, find angles you haven't considered.
Analyze
think through a situation clearly
Break down what's happening. Surface assumptions. See the full picture.
Decide
work through a choice
Weigh options. Find what you actually want. Commit.
Process
sit with how you feel
Describe what happened. No fixing, no advice. Just processing.
Future self
talk to you in five years
The you who got through this is writing back.
Inner critic
steelman the doubts
Give yourself constructive criticism. Sometimes you need to be bad cop with yourself.
Run it as a skill
As a skill
Drop skill/SKILL.md into ~/.openclaw/workspace/skills/mirror/ and your agent picks it up. Works in whatever chat you already use. Recommended.