Integration Guide

This guide helps you choose the right integration type for your use case.
At the core, you decide between Connectors (live integrations), Data Pools (RAG knowledge access), and custom patterns.

Guide: Choosing an Integration Type

1. Clarify requirements

• Do you need real-time data or pre-indexed knowledge?
• Should the assistant only read, or also execute actions?
• Are your systems cloud-based or on-prem/internal?
• Do you need per-user permissions or central admin control?

2. Find the right integration type

• Connectors: For live access and optional write actions in third-party systems
• Data Pools (RAG): For semantic search across large knowledge bases
• Custom MCP: For custom logic, internal APIs, or specialized toolchains
• No-Code: For fast orchestration and automation with minimal custom code

At a Glance

Connectors (AI Tools)

Connectors link your assistant directly with external services like Google Workspace, Microsoft 365, Jira, Confluence, and many others. They work via live API calls and always deliver the most current data.

Advantages

• Real-time data: Always the current state from the source system
• Execute actions: Not just read, but also write (send emails, create appointments, create tickets)
• User permissions: Access corresponds to the individual user's permissions in the source system
• Easy setup: Every user can connect connectors themselves

How it Works

1. The user authorizes once via OAuth with the external service (e.g., Google)
2. When queried, the assistant calls the service's API
3. Data is retrieved and processed in real-time
4. The assistant only gets access to what the user can see in the source system

Typical Use Cases

• Read and write emails (Google Mail, Outlook)
• Manage calendars (view, create, move appointments)
• Work on tickets (create, update Jira issues)
• Cloud documents (current files from Drive/OneDrive)
• Project updates (current status from Asana, Monday.com)

Data Pools (RAG Search Index)

Data pools are pre-indexed document collections that are searched using RAG technology (Retrieval-Augmented Generation). Documents are regularly synchronized and stored in a vector index.

Advantages

• Fast search: Vector search is extremely fast, even with large amounts of data
• Central management: Admins curate knowledge sources for the entire team
• Consistency: All users access the same knowledge base
• Offline-capable: Works even when the source system is unreachable
• Optimized for search: Chunking and embeddings are optimized for text search

How it Works

1. Admins configure the data sources (Confluence, SharePoint, S3, etc.)
2. Documents are automatically indexed and converted into vector embeddings
3. When queried, a semantic search is performed
4. Relevant text passages are provided to the assistant as context

Typical Use Cases

• Search company wiki (Confluence, Notion)
• Policies and handbooks (HR documents, process descriptions)
• Product documentation (technical documentation, FAQs)
• Training materials (onboarding materials, learning resources)
• Archived reports (quarterly reports, analyses)

Activating and Using a Data Pool

Common question: "I selected a data pool, but the assistant cannot find any documents."

This usually happens because the assistant was not saved after selecting the data pool. The configuration is only sent to the backend service when you save.

How to correctly activate a data pool:

1. Open the assistant editor
2. Select the desired pool under "Data Pools"
3. Click "Save" - this step is crucial
4. Start a new chat with the assistant

How does search work in chat?

Once a data pool is active, the assistant automatically searches the knowledge base for relevant questions. You do not need to use any special commands - the assistant recognizes when a search is useful and shows you the found sources as references.

Permissions and Administration

Connectors

• User connects their own account (e.g., Google account)
• Access is limited to the user's data
• No admin setup required
• Ideal for personal productivity assistants

Data Pools

• Central management by IT/admins
• Unified knowledge base for teams or the entire organization
• Control over sensitive company data
• Ideal for knowledge management and support assistants

When to Choose Which Option?

Choose Connectors when you...

• Need real-time data (current emails, calendar events)
• Want to perform actions (create tickets, send emails)
• Need personalized data (only your own files)
• Want to quickly connect something without waiting for admins

Choose Data Pools when you...

• Want to search large document collections
• Want to share company-wide knowledge
• Need consistent answers across all users
• Want to make data from systems without APIs available

Practical Examples

Support Assistant

Recommendation: Data Pool + Connector

• Data Pool: Product documentation, FAQs, known issues
• Connector: Jira for creating and updating tickets

The assistant can find knowledge base articles AND directly create tickets in the ticket system.

Sales Assistant

Recommendation: Connectors

• Google Calendar: Check availability for demos
• CRM Integration: Retrieve current customer data
• Email: Send follow-up emails

Here, real-time data and actions are more important than archived knowledge.

Legal Department Assistant

Recommendation: Data Pool

• Data Pool: Contract templates, compliance guidelines, internal policies

This is primarily about searching documents that rarely change.

Onboarding Assistant

Recommendation: Data Pool + Connector

• Data Pool: Onboarding materials, process descriptions, HR policies
• Connector: Google Workspace for calendar appointments and email notifications

New employees can ask questions AND the assistant can automatically set up appointments.

Combining Both Approaches

The most powerful assistants often use both methods combined:

1. Data Pools for stable company knowledge
2. Connectors for current data and actions

---