Skip to content
AGNT
All playbooks
Onboardingintermediate

Venue Onboarding Playbook

Take a venue from signed contract to live AI agent in one session — scrape, configure, optimize.

AGNT Onboarding Desk3 steps45-60 minutesClaude Sonnet 4.6, Claude Opus 4.6

Why this playbook

Every venue on AGNT needs three things: structured data from its website, a configured agent personality, and optimized booking logic. Doing this manually takes 2-3 hours per venue and introduces inconsistencies across your fleet.

This playbook chains three production prompts that AGNT's own onboarding team runs for every new venue. The output of each step feeds the next — the intake pack.json becomes the hiring brief's knowledge base, and the personality config feeds the booking optimizer's tone calibration.

Run it once to onboard a venue. Run it weekly if you're scaling across multiple cities.

Prerequisites

  • Claude Code installed with filesystem access
  • AGNT MCP server connected (see /stack/mcp)
  • Venue's website URL or a directory of scraped HTML/PDF files
  • Access to AGNT's venue admin panel (for final deployment)

Input requirements

InputTypeRequiredDescription
Venue URL or scraped filesURL or directory pathYesThe venue's public website. If the site is behind a paywall or login, provide pre-scraped HTML/PDF files in a local directory.
Venue categorystringYesRestaurant, bar, cafe, spa, coworking, hotel, or activity. Determines which fields the intake prompt extracts (e.g., "dress code" for restaurants, "check-in time" for hotels).
Booking platformstringNoIf the venue uses OpenTable, Resy, or a custom system, specify it so the booking optimizer can configure the right integration path.

Step-by-step workflow

1

Scrape and structure venue data

Open prompt

Start here. This prompt crawls the venue's website (or reads your scraped files) and produces a structured knowledge pack in AGNT's venue format: JSON for hours, cuisines, and location; markdown for descriptions; and a facts table for policies.

The output is a draft — it marks every field with a confidence score and flags ambiguous data for human review. Never skip the review step.

2

Configure the venue's AI agent

Open prompt

Feed the pack.json from step 1 into this prompt. It generates a complete agent configuration: personality traits, tone of voice, knowledge boundaries, and escalation rules.

The hiring brief enforces AGNT's agent quality standards — it won't let you ship an agent that claims to know things outside its training data or that uses language inconsistent with the venue's brand.

pack.json from step 1
3

Optimize booking and reservation handling

Open prompt

The final step takes the agent config from step 2 and tunes the booking logic: confirmation messages, reminder timing, cancellation policy enforcement, and A2A handoff behavior if the venue uses ClawPulse.

This is where you configure whether the agent books directly or routes through a platform integration.

Agent config from step 2Booking platform details (if any)

Expected outputs

JSON

pack.json

Structured venue knowledge pack with hours, cuisines, location, policies, and confidence scores.

Produced by step 1
YAML

agent-config.yaml

Complete agent personality and knowledge configuration ready for deployment.

Produced by step 2
YAML

booking-config.yaml

Booking flow configuration: confirmations, reminders, cancellation rules, A2A handoff settings.

Produced by step 3

Tool requirements

  • Claude Code with filesystem access
  • AGNT MCP server
  • Venue website URL or scraped files

Troubleshooting

Intake prompt returns low-confidence scores on most fields
The venue's website likely has poor structure or dynamic content. Pre-scrape the site with a headless browser (Playwright) and pass the HTML files directly instead of the URL.
Hiring brief rejects the pack.json as incomplete
Check for missing required fields: venue name, category, at least one contact method, and opening hours. The intake prompt marks missing fields — fill them manually before proceeding.
Booking optimizer produces A2A config but venue doesn't use ClawPulse
Set the booking platform input to "direct" or "manual" to skip A2A configuration. The optimizer defaults to A2A if no platform is specified.
Agent tone doesn't match the venue's brand
Re-run step 2 with explicit tone modifiers in the prompt input: e.g., 'casual and playful for a beach bar' or 'formal and precise for a fine dining restaurant'.

Share as social post

Venue Onboarding Playbook 3 steps, 45-60 minutes. Take a venue from signed contract to live AI agent in one session — scrape, configure, optimize. https://agntdot.com/playbooks/venue-onboarding-playbook

204 / 280 chars

Related playbooks

Run the playbook.

Open each prompt in order, feed the outputs forward, and ship the workflow end-to-end.