Overview
The software supports direct Knowledge Base (KB) retrieval using the Model Context Protocol (MCP). This feature allows approved third-party AI assistants—including ChatGPT, Claude, Gemini, and Atlassian Rovo—to securely search and retrieve official documentation, guides, and help articles.
By exposing our Knowledge Base via an MCP service layer, team members and admins can use their preferred AI environments to ask questions, plan workflows, or automate tasks while receiving accurate, verified answers sourced directly from InflatableOffice’s single source of truth.
How It Works
- User asks a question: You prompt your external AI assistant (e.g., "How do worker availability rules work in InflatableOffice?").
- AI queries the MCP tool: The AI triggers the Knowledge Base MCP tool behind the scenes with your search query.
- Retrieves raw source documentation: The MCP endpoint fetches relevant articles, snippets, and metadata directly from the Knowledge Base index.
- AI crafts the answer: The external AI processes the retrieved documentation using its own language model and displays an answer cited with official software knowledge.
Key Benefits
- Single Source of Truth: Eliminates outdated or hallucinated answers by grounding external AI's in official software documentation.
- Model Agnostic: Works with any MCP-compatible client (ChatGPT, Claude Desktop, Gemini, Rovo, Cursor, etc.).
- Zero Duplication: Reuses existing documentation and indexing without having to sync articles into separate vector databases for each AI tool.
- Cost Efficiency: IO delivers raw search content and metadata without running an intermediate LLM call, letting your third-party tool handle reasoning on its own model subscription.
- Granular Permissions via OAuth: Connections authenticate using standard OAuth linked to individual IO user accounts, ensuring tools only access information the logged-in user is authorized to see.
Security & Access Control
- Internal / Admin Scope: This integration is specifically scoped for internal team members, admins, and workers. It is completely isolated from external or customer-facing AI bots (such as website lead widgets).
- OAuth Authentication: Integrations authenticate via OAuth. API keys are not shared or hardcoded in external client configurations.
- Read-Only Retrieval: The KB MCP tool provides read-only search and retrieval. It cannot create, edit, or delete InflatableOffice records or settings.
Setup & Connection Guide
Prerequisites
- An active account with staff/admin permissions.
- An MCP-compatible AI client (e.g., ChatGPT Plus/Team with custom actions/tools, Claude Desktop, or Rovo).
Step-by-Step Setup
- Open your AI tool’s integration/MCP settings: Navigate to your tool’s external tools or MCP configuration area.
- Select InflatableOffice KB Connector: Enter the MCP endpoint URL (provided in your account settings or staging portal).
- Authorize via OAuth: When prompted, log in with your credentials and authorize the AI client to access KB search on your behalf.
- Verify the connection: Run a test prompt in your AI client:
"Search InflatableOffice Knowledge Base for how to adjust worker availability on desktop."
- Check that the assistant returns accurate snippets referencing official IO documentation.
Example Use Cases & Prompts
|
Workflow |
Example Prompt |
|---|---|
| Feature Deep Dive | "According to the InflatableOffice KB, what are the steps to configure automatic tax rates for multiple counties?" |
| Troubleshooting | "Check IO docs: Why would a customer quote show as pending instead of confirmed?" |
| Drafting Customer Replies | "Using the official IO documentation on delivery fee tiers, help me draft a response explaining how mileage calculation works." |
| Onboarding & Training | "Summarize the IO Knowledge Base best practices for setting up seasonal inventory." |