Skip to main content

Agent Studio

Agent Studio is Orq.ai’s visual interface for building, configuring, and testing AI agents. It provides an intuitive no-code experience for creating autonomous agents that can execute complex workflows, access tools, and manage context without requiring API knowledge.

Overview

Agent Studio allows you to:
  • Build agents visually - Create agents through an intuitive user interface
  • Configure everything - Set up models, tools, knowledge bases, memory stores, evaluators, and safety guardrails.
  • Publish for production - Deploy agents for use via API or Python or NodeJS SDK

Getting Started

Creating Your First Agent

1

Navigate to the Orq Studio

2

Create a new Agent

Use the ’+’ button in your Project or Folder.creating entity
3

Configure your Agent Details

Name and Describe your Agent. You can choose to get assistance using AI to preconfigure agent role and instructions. Choose Start from scratch to configure your agent manually.The Agent Studio editor will open with a preconfigured or default template you can further customize.

Understanding the Interface

The Agent Studio interface consists of two main sections:
  • Instructions Panel (Left) - Define what your agent does and how it behaves
  • Configuration Panel (Right) - Set up models, tools, context, evaluators, and constraints
At any time, save your configuration by using the Publish button.

Configuration Guide

Instructions Section

The instructions section defines your agent’s behavior, goals, and personality. This is where you specify what tasks your agent should perform and how.

Writing Effective Instructions

Use the AI button to help generate effective instructions for your Agent.
Below is an example instruction for your agent. It is recommended to provide exhaustive notes and details on the way the agent should process user input so that their behavior stays as consistent as possible.
You are an experienced customer support specialist for the SaaS company **{company_name}**.  
Your job is to provide clear, concise, and accurate answers to customer inquiries about {product_name}.  
Responses should be brief – no more than **150 words** – and include any necessary next‑step actions.

**Context**  
Customers ask a variety of questions such as how to use a feature, troubleshoot an error, or request billing information.  
You must read the query, identify the main issue, and respond in a friendly tone appropriate for a professional SaaS support environment.

**Step‑by‑Step Instructions**

1. **Read the query** – `{customer_query}`.  
2. **Extract the core problem** (e.g., password reset, API error, pricing).  
3. **Draft a concise answer** – no more than 150 words.  
4. **Add suggested next steps** – at most 3 actions the customer can take.  
5. **End with a friendly closing** and a reminder of available support channels (`{support_contact}`).

**Response Format (plain text only)**

```
Answer:
<brief answer, <=150 words>

Suggested Next Steps:
1. <first action>
2. <second action>
3. <third action>

Thank you for choosing {company_name}. If you need further assistance, feel free to contact us at {support_contact}.
```

**Example**

*Input:*  
`How can I reset my password?`

*Output:*  

```
Answer:
You can reset your password by clicking the "Forgot Password" link on the login page. Enter the email address you used to register, and we’ll send you a secure link to set a new password.

Suggested Next Steps:
1. Go to the login page and click "Forgot Password".
2. Check your inbox (and spam folder) for the reset email.
3. Follow the link and choose a new password that meets our security criteria.

Thank you for choosing {company_name}. If you need further assistance, feel free to contact us at {support_contact}.
```

**Notes**

- Do not include any explanations of the process itself; only provide the answer and next steps.  
- Keep the tone helpful, empathetic, and professional.  
- Use the placeholders exactly as indicated so the prompt can be customized for different customers or products.  

By following these instructions, the AI will consistently deliver short, useful, and brand‑consistent support responses.

Role and Description

Both fields are crucial entries to define your Agent The Description field is used by other agents to understand what an agent can do, describe all the capabilities in details so that multi-agent orchestration behaves successfully when invoking sub-agents. The Role text entry to define your Agent’s responsibility and coverage. This parameter will be sent to Agents during execution to reenforce their perimeter and responsibilities, ensuring their execution fits within their role.
To learn more about multi-agent architecture, see Multiple Agents.

Using Variables

You can reference dynamic values using double braces {{variableName}}. Variables are automatically replaced during execution with actual values passed to the agent.

Model Selection

Choose the language model that powers your agent:
Available models depend on your Model Garden configuration. Models can be switched at any time, and the agent will use the new model on next execution.

Access the model parameters sub-menu to further configure your model.

Considerations when selecting a model:
  • Speed vs Quality - Smaller models are faster but less capable
  • Cost - Larger models cost more per token
  • Capability - Some tasks require more advanced models
  • Latency: Some models use reasoning or thinking tokens, which can add latency to the execution, also see the impact of the Max Iteration and Max Execution Time constraints below.

Tools

Tools extend your agent’s capabilities by allowing it to interact with external systems, execute code, or fetch information. Include tools using the Tool selection modal.
Agent Adding Tool

Browse the tool library to find all tools available in your Project

Standard tools are available to your agents.
ToolNameDescription
Current Datecurrent_dateProvides the Current Date to the Model
Web Searchweb_searchLets an Agent perform a Google Search
Web Scraperweb_scraperLets an Agent Scrape a Web Page
Query Memory Storequery_memory_storeLets an Agent request a Memory Store. Automatically Added when using a Memory Store Context.
Write Memory Storewrite_memory_storeLets an Agent save to a Memory Store. Automatically Added when using a Memory Store Context.
Delete Memory Documentdelete_memory_documentLets an Agent delete a Memory Document. Automatically Added when using a Memory Store Context.
Query Knowledge Basequery_knowledge_baseLets an Agent query a Knowledge Base. Automatically Added when using a Knowledge Base Context.
Retrieve Knowledge Baseretrieve_knowledge_baseLets an Agent fetch a Knowledge Base. Automatically Added when using a Knowledge Base Context.
Call Sub Agentcall_sub_agentLets an Agent invoke another Agent.
You can create your own tools to use within Agents, to learn more see:
Agent model instructions should explicitly mention the available tools to use so that they invoke them.

Context

Add external knowledge sources to your agent:
  1. Click Add context
  2. Select a Knowledge Base or Memory Store.
Knowledge Base Differing from Deployments, here the Knowledge Base is not called with every user query, only when the agent deems it necessary and uses the query_knowledge_base tool. Your Knowledge Base description must be explicit so that the Agent queries it.
To learn more about building your Knowledge Base for Agents, see Knowledge Base.
Memory Store Memory Stores are currently only created and managed through our API. To get correctly used by the agent, the Memory Entity ID needs to be sent during Agent Tasks Runs.
To learn more about using Memory Stores with Agents, see Using Memory Stores with Agents.

Attaching Files

Another type of Context can be attached using files during agent execution.
To learn more about attaching files during agent execution, see Attaching files with Agents.

Evaluators and Guardrails

Evaluators measure agent performance against defined criteria, guardrails can prevent execution on specific evaluations.
Only Pre-configured Evaluators are available to add to agents, to see available standard evaluators and learn how to create custom ones, see Evaluators.
  1. Click Add Evaluator or Add Guardrail
  2. Select evaluator type
  3. Configure evaluation parameters:
    • Define as Input or Output
    • Sample Rate (for Evaluators only)
Evaluators run automatically during task execution and provide performance metrics.
Use multiple evaluators to get comprehensive quality signals. Combine human review with automated metrics for best results.
To learn more about Evaluators and Guardrails, see Evaluators and Guardrails in Deployments.

Runtime Constraints

Control resource usage and execution limits with these constraints. The following controls are available:
ConstraintDescription
Max IterationsMaximum number of tool calls per task
Max Execution TimeMaximum time the agent runs for
Max iteration and Max execution time can have a compounding impact on your agent execution runtime. Start conservative and increase as needed.

Viewing Tasks (coming soon)

The Tasks will let you visualize task executions for your Agent.

Agent and Tasks States

Agents execution can take a long time, if your agent is hanging it is most likely still running execution, please wait and come back to the panel later.
Agents run through the following states when processing tasks:
StateDescription
ActiveExecution in progress, continuation requests blocked
InactiveWaiting for user input or tool results, ready for continuation
ErrorExecution failed, continuation blocked
Approval RequiredTool execution requires manual approval (coming soon)
Tasks go through the following states:
StateDescription
SubmittedTask created and queued for execution
WorkingAgent actively processing
Input RequiredWaiting for user input or tool results
CompletedTask finished successfully
FailedTask encountered an error
CanceledTask was manually canceled

Next Steps

Use the View Code button to see how to integrate your agents in code.
Once your Agent is created, use our SDKs to integrate into your systems. To learn more, see Getting Started with Agents.