Documentation Index

Fetch the complete documentation index at: https://koreai.mintlify.app/llms.txt

Use this file to discover all available pages before exploring further.

​

Tools, knowledge & connections

Studio provides a set of resource management pages for configuring the tools, knowledge bases, external connections, and workflows that power your agents. These resources are shared across all agents in a project and accessible from the Resources and Build sections of the project sidebar.

​

Tools management

Tools extend agent capabilities by connecting them to external APIs, custom code execution environments, and Model Context Protocol (MCP) servers. The Tools page provides a unified interface for creating, configuring, testing, and managing all tool types.

​

Tools page layout

The Tools page displays tools organized by type using a tabbed interface: Each tab shows a count badge indicating the number of tools of that type. A search bar filters tools within the active tab.

​

HTTP tools

HTTP tools allow agents to make REST API calls during conversations. Creating an HTTP tool:

1. Click New Tool and select HTTP Tool.
2. Configure the tool on the creation page:
   • Name — a descriptive name (displayed to the agent).
   • Description — explains what the tool does (used by the LLM to decide when to call it).
   • HTTP method — GET, POST, PUT, PATCH, or DELETE.
   • URL — the endpoint URL. Supports variable interpolation.
   • Headers — key-value pairs for request headers.
   • Authentication — configure API keys, bearer tokens, or other auth methods.
   • Request body — for POST/PUT/PATCH requests, define the body template.
   • Parameters — define input parameters with names, types, descriptions, and whether they are required.

Importing from cURL: You can create an HTTP tool by importing a cURL command. Click New Tool and select Import from cURL, paste a cURL command into the dialog, and Studio parses the command and pre-fills the HTTP configuration (URL, method, headers, body). Review and adjust the generated configuration as needed. Parameter editor: The parameter editor lets you define the input schema for your tool:

• Name — the parameter identifier passed to the API.
• Type — string, number, boolean, array, or object.
• Description — helps the LLM understand how to populate this parameter.
• Required — whether the parameter must be provided.

​

Code sandbox tools

Code sandbox tools execute custom code in an isolated environment during agent conversations. Creating a code tool:

1. Click New Tool and select Code Tool.
2. Configure the sandbox:
   • Name and Description — same purpose as HTTP tools.
   • Runtime — the execution environment (e.g., Node.js, Python).
   • Code — write the tool’s logic in the built-in code editor.
   • Parameters — define input parameters the code receives.
   • Environment variables — set variables available to the code at runtime.

Code tools run in a sandboxed environment with restricted network access and resource limits.

​

MCP servers

MCP (Model Context Protocol) servers provide a standardized way to expose multiple tools through a single connection endpoint. Registering an MCP server:

1. Switch to the MCP tab on the Tools page.
2. Click Register Server.
3. Fill in the registration form:
   • Name — a display name for the server.
   • URL — the MCP server endpoint.
   • Transport type — the connection protocol.
   • Authentication — credentials for connecting to the server.

Each server card shows the server name and URL, connection status, and number of tools exposed. Available actions include View Details (navigate to the server detail page showing all exposed tools), Edit, Test Connection, and Delete.

​

Tool testing

Studio provides an integrated testing interface for verifying tool behavior before connecting tools to agents.

1. Click the test action on any tool card.
2. The test dialog opens with an auto-generated input form based on the tool’s parameter schema.
3. Fill in test values for each parameter.
4. Click Execute to run the tool.
5. Review the results: output data, any error messages, execution latency, and logs from the tool runtime.

The tool detail page also includes a dedicated Test Panel for running tests without leaving the configuration view.

​

Tool actions and connecting to agents

From the tools list, each tool card provides actions for Edit, Preview (read-only view), Test, Duplicate, and Delete. After creating tools, connect them to agents through the agent editor’s Tools section. Add tools from your project’s tool library and configure any parameter mappings or overrides specific to each agent. See Agent development for details on the agent editor.

​

Knowledge bases

Knowledge bases provide agents with access to your organization’s documents and data through semantic search. Studio’s knowledge base management integrates with the SearchAI engine to handle document ingestion, indexing, and retrieval.

​

Knowledge base dashboard

Navigate to Knowledge Bases from the project sidebar to see the knowledge base dashboard. The dashboard displays:

• Summary statistics — total knowledge bases, active (ready) count, and currently indexing count.
• Knowledge base table — a sortable data table listing all knowledge bases with columns for name, status, connector count, document count, and creation date.

Each knowledge base has a status badge:

​

Creating a knowledge base

1. Click Create Knowledge Base on the dashboard.
2. Fill in the creation dialog:
   • Name — a descriptive name for the knowledge base.
   • Description (optional) — explains the purpose and content scope.
   • Language — the primary language of the documents.
3. Click Create.

The knowledge base is created in “creating” status and transitions to “ready” once the initial index build completes.

​

Knowledge base detail page

Click any knowledge base in the dashboard table to open its detail page, organized into the following tabs: Overview tab — shows knowledge base metadata, status, and configuration summary. Includes statistics on document count, chunk count, and index health. Documents tab — manages the content in your knowledge base. Upload files (PDF, DOCX, TXT, HTML, Markdown) that are processed through the ingestion pipeline: parsing, chunking, embedding generation, and indexing. View document metadata, preview content and chunks, and delete documents to remove them from the index. Connectors tab — enables automatic ingestion from external data sources. Add connectors through a wizard, set up synchronization schedules, view connector status and last sync time, filter by status, and trigger manual synchronization. Web crawler tab — indexes content from websites. Configure a start URL, crawl depth, URL include/exclude patterns, and custom domain lists. Monitor crawl job history with status, page count, and duration, and preview crawled pages with their extracted content. Fields tab — configure custom metadata fields for documents in the knowledge base. Fields can be used for filtering search results. Vocabulary tab — define custom vocabulary entries for domain-specific terms with definitions and synonyms. The vocabulary helps the search engine understand specialized terminology. Test vocabulary recognition with the vocabulary test panel. Knowledge graph tab — configure entity extraction and relationship mapping across your documents. Enable knowledge graph features, select extraction models, configure taxonomy categories, and set up workspace-level inheritance for shared taxonomies. Query playground tab — test search queries against the knowledge base. Enter a natural language query, view search results with relevance scores, inspect individual chunks, and adjust query pipeline settings. Settings tab — configure model selection for embeddings and query processing, chunking strategy, index rebuild controls, and knowledge base deletion.

​

Connecting knowledge bases to agents

After creating and populating a knowledge base, connect it to agents so they can use it during conversations. Knowledge base connections are managed through the agent’s tool and configuration settings. See Agent development for details.

​

Connections & integrations

Connections enable your project to communicate with third-party services such as CRM systems, messaging platforms, ticketing tools, analytics services, and data sources. Studio provides a connector catalog and connection management interface.

​

Connections page layout

The Connections page is organized into two sections: My connections — the top section displays your active connections as compact cards showing the connector logo, connection name, a status indicator (green for connected, red for error), and last verified timestamp. Click a connection card to expand an inline detail panel with configuration details and available actions. Connector catalog — the bottom section displays the full catalog of available connectors, organized by category: Each catalog card shows the connector name, logo, a brief description, and a Connect button.

​

Creating a connection

1. Find the connector in the catalog and click Connect (or click Add Connection from the My Connections section).
2. In the creation modal, fill in:
   • Connection name — a descriptive label for this connection instance.
   • Credentials — service-specific configuration (API keys, OAuth tokens, endpoint URLs, etc.).
   • Options — any connector-specific settings.

For connectors that use OAuth, Studio opens an authorization window for the target service. Sign in and grant permissions, and on successful authorization you are redirected back to Studio with the connection established. The OAuth flow dialog manages the token exchange process and stores refresh tokens securely.

​

Testing and managing connections

After creating a connection, click on the connection card and use the Test button to verify connectivity. Studio sends a test request to the external service and shows diagnostic information on failure. From the expanded connection panel you can also Edit connection details (name, credentials, options) or Delete the connection. A Connection Status Bar provides an at-a-glance summary of all connection statuses across your project, showing counts of healthy, warning, and error connections.

Tip: Test connections periodically, especially after credential rotations or service configuration changes.

Warning: Deleting a connection may break agents or workflows that depend on it. Review the connection’s usage before deleting.

​

Using connections in agents

Connections are used by agents through tools. When you create a tool that calls an external API, you can associate it with a connection to use managed credentials rather than hardcoding authentication details.

​

Workflows

Workflows define multi-step processes that coordinate agent actions, external service calls, and human approvals. Use workflows to automate business processes that span multiple systems or require structured orchestration beyond a single agent conversation.

​

Workflows list

Navigate to Workflows from the project sidebar to view all workflows in your project. The list page provides a search bar, status filter (Active, Archived, Paused), a Create Workflow button, and workflow cards in a two-column grid showing name, description, status badge, step count, and last-modified timestamp.

​

Creating a workflow

1. Click the Create Workflow button.
2. In the creation modal, fill in:
   • Name — a descriptive name for the workflow.
   • Description (optional) — explains what the workflow automates.
   • Trigger type — how the workflow is initiated (manual, scheduled, event-driven, or agent-triggered).
3. Click Create to be taken to the workflow detail page.

​

Workflow detail page

The workflow detail page provides a comprehensive view of the workflow’s configuration, execution steps, and run history. Workflow steps — define the sequence of actions:

• Agent steps — invoke an agent to handle part of the process.
• Tool steps — execute a tool (HTTP call, code execution).
• Approval steps — pause execution and wait for human approval.
• Conditional steps — branch based on data or previous step results.
• Delay steps — wait for a specified duration before continuing.

Status management — workflows have a lifecycle status of Active (operational and can be triggered), Paused (temporarily disabled), or Archived (no longer in use but retained for reference). Change status from the detail page actions menu.

​

Execution monitoring

The workflow detail page shows a history of all executions (runs). Each run displays a run ID, status (succeeded, failed, running, or waiting), start time and duration, and expandable step results. For running workflows, you can monitor step-by-step progress in real time with timing information for each step.

​

Workflow approvals

When a workflow reaches an approval step, a task is created in the Inbox (see Testing, deployment & operations). The workflow pauses until a team member reviews and approves or rejects the step.

​

Related pages

• Agent development — building agents that use tools and knowledge bases
• Studio overview — navigation and project management
• Testing, deployment & operations — deploying and monitoring agents
• Insights & settings — guardrails and project configuration