Quickstart Guide
Getting Started with Chicory
Follow this guide to sign up for Chicory AI, link your data, train your model, and create your first AI agent. These steps will help you quickly experience the main features of Chicory AI.
Sign Up & Access Your Chicory AI Workspace
Open your browser and go to
app.chicory.ai
.New users: Click "Sign Up." Provide your name, email and create a strong password.
After verifying your email, you'll be prompted to create an organization by entering its name. This will be your workspace.
Once your organization is set up, you can invite other members. This gives team members shared access to context and AI agents, promoting collaboration. You can do this right away or later in your organization's settings
Complete any remaining steps to finalize your account and organization setup.
Chicory Signup Screen Existing users: Log in with your credentials.
Connect Your Assets in the "Integrations" Tab
To understand your data well, Chicory AI needs to connect to your company's various data assets. You'll manage these connections in the "Integrations" tab.
How to Add an Integration:
Open the Integrations Tab
In the left-hand menu of the dashboard, click Integrations.
Pick Your Asset Type
You’ll see three sections: Document Sources, Code Repositories, and Data Sources.
Click the section for the asset you want to add.
Select Your Connector
Choose from our list (e.g., Databricks, GitHub, Google Drive, file upload).
Enter Connection Details
Test & Save
If there’s a Test Connection button, click it to confirm everything works.
Then click Save or Connect.

Tip for Best Results: To get the most out of Chicory AI, help it understand your data fully. This means connecting:
Your main Data sources (like databases or data files).
Your Code (like GitHub repositories) that creates, changes or uses your data. Make sure the access token has access to the repositories you plan to train on!
Your Docs (like PDFs, Word files, or items from Google Drive) that explain your data, business rules, or how things work.
Connecting all three types of assets (Data, Code, and Docs) for the same project helps Chicory AI understand things much better.
Train Chicory
Start Training
Once all data sources are added, it's time to train Chicory on your unique data landscape. Find the Train button located on the same Integrations tab, and click to begin training.

What happens
Chicory scans schemas, parses code, indexes docs, samples the data and links everything into a living "context". For how it works, refer to Understanding Chicory.
Training Duration and Support
Training time varies based on the volume, type, and complexity of the connected assets
Important: If Chicory's training has not completed within 1 hour, or if processing very large assets, contact [email protected]
for immediate assistance.
Create and Test your first AI Agent
Start creating an Agent: In the main navigation panel, click the "+ Agent" button.
Define Your Agent:
Name: Enter a clear name, e.g., "Data Catalog Assistant."
Description: Describe its purpose, e.g., "Provides descriptions for tables and columns within a specified connected data source when requested by a user."
Instructions: Provide natural language instructions that define the agent's core capabilities and how it should respond.
Example:
You are a Data Catalog Assistant. Your main role is to answer user questions about the details of tables and columns within this data source.This includes providing summaries of tables, listing their columns, and giving column data types and any available descriptions that Chicory AI has discovered. Always aim to provide clear, accurate, and helpful information.
Output Format (if applicable): While the initial interaction is chat-based, the agent should aim for structured and readable responses within the chat that can be used by future API, and Agent to Agent communications. Please see step 5 for details on how to deploy agents.
Hit "Create Agent" and your first Chicory-powered agent is ready to go!
Test & Refine in Chat:
After saving your agent, use its dedicated chat interface to interactively test and refine its behavior. Engage by typing natural language requests or commands that align with the instructions you set during its definition.
For example, if you configured the Data Catalog Assistant, you might ask it:
"Describe the 'customer_activity' table."
or"What columns are in the 'product_inventory' table?"
Carefully review the agent's responses to your inputs. You can then ask follow-up questions, use different phrasings for your requests, or test various scenarios to ensure the agent performs correctly.
Review the agent's responses in the chat. Test its performance by asking follow-up questions, rephrasing requests, and exploring different scenarios. If adjustments are needed, click the agent's name and edit its instructions. Save changes, then return to the chat to re-test and refine.
Deploy your Agent as ACP (A2A coming soon)
Click the Agent’s Name: At the top of the chat click on Agent's Name
Click "Edit": This opens the agent’s configuration.
Find the "Deployed API" Section: Navigate to the section labeled “Deployed API”
Deploy the Agent: Click “Deploy Agent”
View Deployment Info: After deployment, you’ll see the API endpoint, authentication details, and version info.
Last updated