Use Cases and Playbooks
MCP servers connect the tools your company built in Assist to the AI client you already use. These playbooks walk you through building real systems -- tools, apps, and agents -- entirely from inside Claude Desktop, ChatGPT, or another MCP-compatible client. You do not need to log into Assist to build any of this.
Each playbook follows the same pattern: connect to your company's systems, explore the data conversationally, build tools that wrap the safe operations, then build a sandcastle app and agents that operationalize the workflow for your whole team.
Playbooks
Build an Order Management App on Top of Your WMS
Your warehouse management system has the data everyone needs but only a few people can access. Build tools that connect to the WMS API, then build a sandcastle app where the ops team can search orders, track fulfillment status, and flag exceptions -- all with persistent state that survives between sessions.
Build a Campaign Launcher for Your Internal Ad Platform
Your demand-side platform requires ad ops to configure every campaign. Build tools with built-in guardrails (budget caps, required fields, validation), then build a sandcastle app where account managers configure and submit campaigns themselves. Ad ops reviews and approves instead of doing every launch manually.
Build a Supply Chain Monitoring Agent
Someone on your procurement team spends an hour every morning checking inventory against demand. Build an agent that runs the check automatically, compares today's risks to yesterday's, and alerts the team in Slack when something needs attention. Build a sandcastle dashboard where the team tracks risks, logs actions, and monitors trends over time.
Build an Intake App for a System Only Engineers Can Use
Your company has an internal system (feature flags, deployment pipeline, data warehouse) that non-engineers need to use but cannot access. Build guarded tools that expose the safe operations, then build a sandcastle app with forms, validation, and an approval workflow. Non-engineers serve themselves for routine tasks. Engineers review the risky ones.
Build a Reconciliation Agent Across Two Systems
Your finance team manually reconciles data between two systems every month. Build an agent that runs the reconciliation weekly, tracks discrepancies over time with persistent state, and alerts the team to new mismatches. Build a sandcastle app where the team investigates discrepancies, assigns work, and maintains an audit trail.
Build a Client Reporting Portal from Your Analytics System
Generating client reports requires an analyst to log in, run queries, and format the output. Build query tools with safe parameterized access, then build a sandcastle app where account managers generate formatted reports with charts and AI-generated takeaways. Add templates for recurring reports and a scheduled agent that generates them automatically.