Own your
knowledge.
Every AI model you use shares one permanent memory. Switch from Claude to GPT to Grok without losing context. Your knowledge. Your control. Zero lock-in.
Your memory is infrastructure. Not a feature.
Every model you use forgets you tomorrow. mybrain is a durable memory layer — queryable, exportable, portable — that every model reads from and writes back to.
Compounds automatically
mybrain only learns what's genuinely new. Every conversation adds to a growing graph — no duplicates, no noise.
Portable across models
GPT, Claude, Gemini, Grok, Llama, Mistral, DeepSeek. Your memory and context travel with you on every model switch.
Queryable via REST
Any agent — Codex, Claude Code, your own scripts — can read and write memories through one endpoint.
Isolated by design
Your vault is scoped to you alone. Isolation is enforced at every layer of the stack.
Obsidian-compatible
Your brain lives as structured markdown in Cloudflare R2. Export anytime. No lock-in.
No training, ever
We don't learn from your data. Your conversations never touch a training pipeline. Period.
It doesn't just remember. It acts on your behalf.
Connect Gmail and Google Calendar and mybrain becomes a second set of hands. It reads threads that need attention, drafts replies in your voice, and negotiates meetings — with your memory as context and your sign-off as the guardrail.
Read
Ingests your inbox and calendar on a schedule. Surfaces what needs you — drops what doesn't.
Write
Drafts replies in your voice, grounded in your memory and the full thread context.
Schedule
Finds a slot that works across calendars, pulls buffers from your preferences, sends the invite.
Respect
Nothing auto-sends without your sign-off by default. Every action is reviewable and undoable.
Value compounds. Slowly. Then all at once.
Day 1 you describe yourself. Month 6 it describes you better than you do. The graph gets sharper on its own — synthesis, gap-detection, key-truth elevation run in the background.
Two ways to run your brain.
Plus is managed — we host the memory layer and bundle every model. BYO plugs your API keys in and bills you for memory only.
| inference | your API keys |
| models | Anthropic, OpenAI, OpenRouter |
| memory vault | unlimited |
| documents | unlimited |
| api calls | not included |
| webhooks | — |
| support |
| inference | included, all models |
| models | every release + betas |
| memory vault | unlimited |
| documents | unlimited |
| developer api | included |
| webhooks | included |
| support | priority |
// 7-day free trial · cancel anytime · no per-message fees
Your brain. Every app.
Any agent that speaks HTTP can read and write your memory. One API key, user-scoped, rate-limited, auditable.
# Ask your brain anything — from any app
curl -X POST https://api.mybrain.ai/v1/brain/ask \
-H "Authorization: Bearer mbk_live_..." \
-H "Content-Type: application/json" \
-d '{"question": "What are my dietary preferences?"}'
# Response
{
"answer": "You prefer vegetarian. Allergic to peanuts.
Love Thai (from 47 past conversations).",
"confidence": "high",
"sources": ["preferences/food.md", "conversations"]
}