What it includes
The November 2025 changelog has the full details.
Agents interface
You create AI agents in the dashboard. For each agent, you choose the model, the system prompt, the tools and the vector database collections that it uses.
Chat panel for testing
You can test an agent in a chat panel in the dashboard, and check how it answers your questions.
AI tool builder
The tool builder lets you create and edit custom tools and their edge function logic. Each custom tool runs as an edge function. When you deploy a tool, the builder applies the timeout, category, response mode and async settings that you enter.
Vector database collections
A vector database interface, also new this month, lets you create and search vector collections in the dashboard. You can give an agent the collections that it needs.
API endpoints and SDKs
New API endpoints expose tools and tool executions for programmatic use. The TypeScript, PHP and Go clients at v4.4.0 include the new AI tools and vector database endpoints.
Who it is for
Teams building assistants
Set up an agent with its own model, system prompt and tools, and then test its answers in the dashboard chat panel.
Developers
Write custom tools as edge function logic in the tool builder, and work with tools through the new API endpoints.
Content teams
Create and search vector database collections in the dashboard, and choose the collections that each agent uses.
What it is
AI agents are part of the Quant AI platform. An agent combines four things that you choose in the dashboard: a model, a system prompt, a set of tools and a set of vector database collections. The system prompt tells the agent what its job is. The tools give it more things that it can do. The collections give it content that it can search.
The AI tool builder, which is also new this month, is where you make your own tools. You write the edge function logic for each custom tool, and the tool runs as an edge function. Agents, tools and vector collections are all in the dashboard, next to the rest of your AI tooling.
How it works
Take an agent that answers questions about your own documentation as an example.
First, you create a collection for the content. In the dashboard sidebar, select Vector Database, and then select Create Collection. Give the collection a name and a description, and choose an embedding model. You can then search the collection from the dashboard.
Next, if the agent must do more than answer from your content, you create a tool. In the tool builder, write the edge function logic for the tool. In the form, set the timeout, category, response mode and async settings, and then deploy the tool.
Then you create the agent:
- Select AI Agents in the dashboard sidebar, and then select Create Agent.
- Enter a name and a description.
- Select a model and write the system prompt. For example, tell the agent to answer only from the documentation collection.
- Assign the custom tool and the vector database collection.
- Save the agent.
Last, you test the agent in the chat panel. Ask the questions that your users ask, and check the answers. If an answer is not correct, change the system prompt, the tools or the collections, and test again.
For programmatic use, the new API endpoints expose tools and tool executions, and the v4.4.0 SDK clients include them.
Get started
Open AI Agents
In the dashboard sidebar, select AI Agents, and then select Create Agent.
Name the agent
Enter a name and a description that say what the agent is for.
Choose model and prompt
Select an AI model, and then write the system prompt for the agent.
Add tools and knowledge
Assign tools and vector database collections, and then save the agent.
Test in the chat panel
Ask the agent questions. Read Creating your first agent.
Related: Explore AI agents · Documentation · November 2025 changelog
Ready to try it on Quant?
Use AI agents on your own project, or talk to our team first.