studio-web
2.2510
true
Studio Web user guide
- Getting started
- Administration
- RPA workflow projects
- Creating an RPA workflow from an idea
- Creating a project
- How to start an RPA workflow
- Managing project files and folders
- Connecting RPA workflows to your accounts
- Configuring activities
- Managing the activities in a project
- Passing values between activities
- Iterating through items
- Managing the data in a project
- Configuring a project to use your data
- Using file and folder resources
- App projects
- Agentic processes
- Agents
- Solutions
- API workflows
- Tests
Snowflake Cortex Agents unlock natural language interaction with your data, helping you query, analyze, and act on enterprise datasets.
This workflow acts as an intermediary to run an "ORDERS_AGENT" in Snowflake. It takes a question as input, sends it to the Snowflake Cortex Agent API, and then returns the API response directly as the workflow output.
Prerequisites
First, you need to have a Cortex Agent set up in your Snowflake account.
- Create an API workflow.
- Define workflow inputs. Open Data manager and add a new Inputs parameter with the following configuration:
- Name—Question
- Type—String
- Leave the Required checkbox empty.
- Add a HTTP activity and configure it as follows:
- Method—POST
- Request URL—Provide the Cortex Agent run endpoint, specific to your Snowflake account and agent, for example
"https://{your_snowflake_env}/api/v2/databases/SNOWFLAKE_INTELLIGENCE/schemas/AGENTS/agents/ORDERS_AGENT:run" - Headers—
{ "Content-Type": "application/json", "Accept": "application/json", "Authorization": "Bearer <Add your snowflake_token here>" }{ "Content-Type": "application/json", "Accept": "application/json", "Authorization": "Bearer <Add your snowflake_token here>" } - Request body—
{ "messages": [ { "role": "user", "content": [ { "type": "text", "text": $workflow.input.Question } ] } ], "toolChoice": { "type": "auto" } }{ "messages": [ { "role": "user", "content": [ { "type": "text", "text": $workflow.input.Question } ] } ], "toolChoice": { "type": "auto" } }
- Add a Response activity and leave it as is. This captures and returns the output from the Cortex agent as a JSON.
- Run the integration. Once deployed, you can invoke the workflow with any natural language query. The workflow passes the question to the Cortex Agent, receives the streaming output, and returns a structured JSON response that can be used in downstream automations.
Key take-aways
- Streaming is seamless – Although Cortex API is streaming-only, the API workflow built-in support meant no additional coding was required to parse and collect results.
- Consistent API contract – All Cortex Agents share the same request/response shape. Your integration does not need to change as your agents become more advanced.
- Rapid prototyping – From setup to working integration took minutes. This speed allows teams to quickly validate use cases and iterate.