An AI assistant is a program that uses a large language model (LLM) to reason and take actions on your behalf. Unlike a simple chatbot that only responds to questions, an assistant can plan a sequence of steps, use tools, read and write data, and complete tasks end-to-end with minimal human input.
How assistants work
- You give the assistant a goal or a trigger (for example, "summarise all new support tickets each morning")
- The assistant breaks the goal into steps using its reasoning capabilities
- It uses connected tools — APIs, databases, integrations — to carry out each step
- It returns a result or takes an action (sends a Slack message, updates a spreadsheet, creates a ticket, etc.)
What assistants can connect to
On QuivaWorks, assistants can be connected to integrations including:
- Communication — Slack, email
- Productivity — Google Drive, Notion, Microsoft 365
- CRM & support — HubSpot, Zendesk, Salesforce
- Payments — Stripe
- Custom APIs — via the MCP framework
Assistants vs automations
Traditional automations follow fixed rules: if X happens, do Y. AI assistants are more flexible — they can handle ambiguous instructions, adapt to varying inputs, and make judgment calls that rule-based tools cannot.
Learn more
MCP (Model Context Protocol) is an open standard that defines how AI models connect to external tools and data sources. It provides a consistent way for assistants to call tools — whether that's querying a database, triggering an API, or reading a file — so assistants built on different models can use the same tools reliably.
Why MCP matters
Before MCP, every AI integration had to be built custom for each model and each tool. MCP standardises the interface, so:
- A tool built once works with any MCP-compatible assistant
- Assistants can discover what tools are available and how to use them at runtime
- Tool developers publish once to the Marketplace rather than building separate integrations for every platform
MCP tools on QuivaWorks
The QuivaWorks Marketplace includes a growing library of MCP tools. Each listing shows:
- What the tool does and what data it can access
- What permissions it requires
- Pricing (free or paid)
- How to connect it to an assistant
QuivaWorks also includes a utility that can convert any API into an MCP-hosted tool, so you can wrap your own internal APIs without writing an integration from scratch.
For developers
If you want to build and publish your own MCP tool, see the MCP developer guide in the documentation.
Browse MCP tools
Visit the Marketplace and filter by MCP Tools to see what's available.
Installing an assistant from the Marketplace is straightforward and doesn't require any coding.
Steps
- Go to the Marketplace
- Use the search bar or browse by category to find an assistant
- Click on the listing to open it — read the description, required permissions, and pricing
- Click Deploy to workspace
- Select your workspace from the dropdown
- You'll be shown a list of integrations the assistant needs — connect each one using the authorisation flow
- Fill in any configuration options (these vary by assistant)
- Click Activate
The assistant will appear in your workspace dashboard under Active Assistants.
What if I need to connect an integration I haven't used before?
When you install an assistant that requires a new integration, you'll be prompted to authorise it. This uses standard OAuth where available — you'll log into the third-party service and grant the permissions the assistant needs.
You can also pre-connect integrations from Workspace Settings → Integrations before installing an assistant.
Can I install the same assistant in multiple workspaces?
Yes. You can deploy an assistant to any workspace you have access to. Pricing and usage credits are tracked per workspace.
Uninstalling an assistant
Go to your workspace dashboard, find the assistant under Active Assistants, and click Deactivate. This stops the assistant from running but doesn't delete its configuration. You can re-activate it later from the same screen.