Grok-Bot Grillo Setup

Hosted connector for independent runtime behavioral attestation on xAI Grok Bot

Hosted connectorT0-GB

What this is

Grillo on Grok Bot quizzes another teammate, sends answers here for scoring, and returns a verify URL. Completions stay on Grok Bot; AI Assess Tech is score-only. This is attestation — not a management-system badge and not a self-certifying VALIDATED stamp.

Three products — pick one

Hosted Grok-Bot connector

This page

Grillo on xAI Grok Bot calls our hosted MCP. Interview stays on Grok Bot; platform is score-only.

You are here

Customer-hosted MCP

T0-CH

Run @aiassesstech/grillo-mcp on your machine with a tunnel. Different Authorization secret.

npm README

OpenClaw fleet Grillo

VPS fleet

Six-agent OpenClaw VPS fleet plugin — different install and product.

OpenClaw docs

What it does

✓Runs the 120-question LCSH interview on Grok Bot in packs (default pack size 15)
✓Scores answers on AI Assess Tech and returns a public verify URL
✓Supports dry-run preflight (5 questions, mock scores, no billing)
✓Supports resume if the Grok-Bot host drops mid-interview
✓Reports drift in results — does not take a Grok-Bot teammate offline

What it does not do

✗No posting or fixing — attestation only; Grillo does not remediate the subject
✗No self-assessment — never attest grillo (any case). Jarvis names the subject teammate
✗No VALIDATED or ISO claims — independent runtime behavioral attestation only
✗No OpenClaw install — do not mix fleet plugin steps into this connector path
✗No second Grillo teammate — use the existing Grillo Grok-Bot assessor seat

Public Grok Bot template: live at https://x.ai/bot/0K8ZjhfS8SUgIRKt9hS0T. Connector steps below still apply.

Operator setup

Mint the Assess Key

Sign in at aiassesstech.com
Open Settings → Assess Keys
Generate a new key with an obvious name (e.g. Grok-Bot Grillo)
Copy hck_… immediately — it is not shown again
On the success card, Connect Grillo on Grok-Bot shows the Connector URL and Authorization fields

Connector URL (production): https://www.aiassesstech.com/api/mcp/grillo

See also: Settings guide — Assess Keys

Paste into Grok-Bot

In xAI: Settings → Plugins and/or grok.com Custom connector
Add a remote MCP connector with the URL from Settings and Authorization = your Assess Key
Do not paste an Agent API Key, provider key, or self-hosted npm secret
Do not create a second Grok-Bot named Grillo

Jarvis names the subject

Jarvis (Chief of Staff) names who to attest — for example, a teammate like Voss. The assessor seat (Grillo) is never the subject. Never attest grillo in any case.

Preflight — dry run (no bill)

grillo_setup — check, validate, guide
grillo_bank with dryRun=true — expect 5 questions, pack size 15, and a minted sessionId
Interview the subject; submit the letter the teammate said (A–D)
grillo_submit with dry run + finalize=true — mock scores only

If preflight fails, stop. Do not start a 120-question live interview.

Live interview

grillo_bank (no dry run) — keep sessionId
Work in packs of 15 (~8 submit hops for 120 questions)
Send question text and options exactly as returned; submit the letter the teammate said
After each pack: grillo_submit with finalize=false
Last pack: grillo_submit with finalize=true

Resume if the host dies

grillo_status — recent runs and verify URLs
grillo_bank with resumeRunId + same sessionId
grillo_submit to continue — same verify URL

Read the result

Open the verify URL from submit or status. Use attestation language in chat and reports.

Hosted connector tools

Tool	Purpose
grillo_setup	Preflight: check, validate, guide
grillo_bank	Fetch bank; mint or resume sessionId
grillo_submit	Submit answer letters per pack; finalize at end
grillo_status	List recent runs and verify URLs