Your Identity
Context for AI.
Write who you are once. Every AI agent starts already knowing you — your background, your projects, your decisions. And as you work, agents write back what they learn, so your Meishi grows with you.
Free to start. No credit card.
Every new AI. Same introduction.
Claude
"I'm a PM building a SaaS tool. Our stack is Next.js and Postgres. We've decided to..."
Cursor
"I'm a PM building a SaaS tool. Our stack is Next.js and Postgres. We've decided to..."
ChatGPT
"I'm a PM building a SaaS tool. Our stack is Next.js and Postgres. We've decided to..."
You paste your background. You explain your stack. You repeat your decisions.
Every. Single. Session.
AI tools have no memory between sessions, no shared understanding across tools, and no way to know what you've already decided. You become the integration layer — manually re-briefing every agent, every time.
How it works
Three steps. Then never again.
01
Write once
Fill in who you are, what you're building, and what you've already decided. Plain text.
02
Connect an agent
Paste one snippet into any agent's system prompt. That's it.
03
They remember
Every agent starts informed. Decisions accumulate. Context compounds.
The API
What agents see
On session start, an agent calls GET /api/v1/context and receives everything you've written — your role, current projects, past decisions, things not to repeat, and every note agents have left over time.
Agents can write back with /write or leave observations with /append. Every change is logged, attributed, and reversible.
// every agent reads this on session start
{
}
Built right
Nothing you have to think about.
AES-256 encrypted
Your context never touches our DB as plaintext. Unique key per user.
Full audit log
Every read and write is recorded — who, when, before/after.
Version history
Snapshots before every agent read. Restore any version instantly.
Scoped API keys
Read-only, write-only, or full access. Revoke any key instantly.
Works with any agent
Plain Bearer token REST API. Claude, GPT, Cursor — all of them.
Append-only notes
Agents leave timestamped notes. You read them. Nobody overwrites them.
FAQ