MindClaw

Research Preview

Train a LoRA for every OpenClaw on every conversation for personalized long-term memory.
Preparing your workspace…
Checking authentication, agreement state, and runtime health.
Three Surfaces

Talk. Watch. Collect.

This alpha routes each signed-in user to a dedicated MindClaw runtime. Chat directly, watch MinT training activity, and track checkpoint snapshots from the same workspace.

Chat Keep one live session or rotate to a new one whenever you want a fresh training thread.
Training Logs See rollout and trainer activity as your runtime learns from recent turns.
Checkpoints Track checkpoint metadata as MinT produces new resumable states.
Account

Sign in with email

Use your account email. New users will verify by email and set a password during the first login.

Registered accounts sign in with email + password. Unregistered or password-unset accounts require an email verification code.
Experimental Agreement

Read before entering

MindClaw is a Research Preview. Please review the User Agreement and Privacy Policy before entering your workspace.

Starting Model

Choose your starting MinT model before the workspace is provisioned. We recommend staying on Auto unless you specifically need a larger model.

Switching models later may clear training progress and checkpoints.
Capacity Notice

Sorry, we are overloaded.

Signup to get email notification when we add more resources.

Your runtime is not available right now. We keep your account and agreement state, and you can retry after more capacity is added.

Workspace Loading…
Idle
MinT Checking
Model: Auto (qwen 4b)
session: pending

Chat

Each message flows into your dedicated MindClaw runtime.

web-session
Press Ctrl/Cmd + Enter to send.

Training Logs

Rollout, trainer, and sync activity from your user container.

0 items

Checkpoints

MinT-backed checkpoint metadata captured from your MindClaw trainer.

0 items