PublicMITGrok-only

X DM safety Bot × SlimeTree-RLM

▸ Don't send brand-account DM spam to your AI

Inbound DMs are classified D/μ/R by the RLM, then the bot replies automatically. Investment solicitations / affiliate spam are mechanically suppressed at μ; only genuine inquiries are delegated to Grok at R.

1. X DM API credentials

The X DM API requires X Pro tier ($5,000/mo) or higher. Issue a User Access Token (scopes dm.read + dm.write) at developer.x.com.

X User Access Token

recipient User ID

2. Grok (xAI) API key

Grok key

Platform-native LLM: X DM processing is Grok-only.

2.5 Operating mode ★ NEW: B (within-Grok tier) ― shared module

Single LLM ― the single Grok above Cost-tier (B) ― cheap → judge → escalate only the insufficient to premium

3. Inbound DM simulator

Play the "customer" role and type a message → the bot replies via the RLM. Examples: What's the price? → D / a → μ / compare 3 industry case studies → R

D: 0

µ: 0

R: 0

Send: 0

cheap: 0

esc: 0

esc rate: ―

Send a "customer" message from the form below.

4. API log

(not yet run)

5. Effects

Auto-block DM spam: investment solicitation / affiliate / phishing are mechanically suppressed at μ
D instant: hours / product specs answered 0-token, 24h
R Grok: only borderline cases are delegated to the X-native LLM
WAL chain: for harassment cases, "when did what arrive and how did the bot handle it" can be reproduced air-gapped