Skip to main content

AI Service Panel

The AI Service panel gives you visibility into the Tools, Knowledgebases, Agents, Teams, Prompt Templates, and External MCP Servers in your Twinit projects.

AI Service Panel Project Folder Commands​

Configure MCP​

The Configure MCP command aides you in configuring MCP connections to your Twinit workspaces.

The Configure MCP form guides you through generating your MCP API keys, selecting the MCP tools in the workspace to use, and generating the JSON mcp servers config.

If using Cursor, the configuration can be automatically brought over to the Cursor settings and installed.

Agents Folder Commands​

Create Agent​

Allows for creating and configuring new Twinit AI Agents by providing an Agent name, background, user type, and an AI provider and model.

Optionally, a description can be provided and tools, knowledgebases, and a Prompt Template can be configured for the Agent.

To assign a Prompt Template:

  1. Open Create Agent from the Agents folder
  2. In Prompt Template, choose one template from the project or leave None
  3. Templates are listed as fixed (no arguments) or dynamic (placeholders filled at runtime from conversation context)
  4. Click Save Agent

The selected template is attached to the agent and injected as its instruction when a team runs. Choose None to create the agent without a template.

Optionally enter an Invocation Message (maximum 500 characters). When this agent starts working in a streaming Chat with Team conversation, that message is shown in the activity log. Leave it blank to show no custom message.

External MCP Servers can also be enabled for the Agent, optionally filtering the tools exposed to the Agent from each server.

Agent Commands​

Delete​

Deletes the Agent from the AI Service.

Edit Agent​

Opens the Agent allowing for changes to be made to its configuration, including the assigned Prompt Template, Invocation Message, the External MCP Servers enabled for the Agent, and the tools exposed from each server.

Change Prompt Template to assign a different template, or choose None to remove the assignment from the agent. Clear Invocation Message to stop showing a custom line when the agent is invoked in streaming chat.

View Agent JSON​

Opens the full Agent JSON Twinit configuration in a new file.

Custom MCP Tools Folder Commands​

Create Custom MCP Tool​

Opens a form to configure a custom MCP tool in the workspace.

To configure a tool:

  1. Give the tool a clear meaningful name an agent can use to understand the tool's purpose
  2. Enter a detailed description including what the tool does, when and how it should be used, and any other information to inform the agent's usage of the tool
  3. Select the script on Twinit to use to back the tool (this must be a backend script, not a client script with imports)
  4. Select the function in the script to use to back the tool
  5. Create a schema (with field descriptions) for the tool to use to validate any input the tool is given
  6. Click Create

After creating a new MCP Tool, you may need to restart your MCP Server Connection or your IDE.

Custom MCP Tool Commands​

Edit Custom MCP Tool​

Edit an existing Custom MCP Tool.

If using Cursor, the newly edited configuration can be installed via settings and the previous version will be automatically overwritten.

After editing an MCP Tool, you may need to restart your MCP Server Connection or your IDE.

View Custom MCP Tool JSON​

Opens the full Custom MCP Tool JSON Twinit configuration in a new file.

Delete​

Deletes a Custom MCP Tool configuration.

External MCP Servers Folder Commands​

The External MCP Servers folder allows you to connect external Streamable HTTP MCP servers to your project. Expand a server in the tree to see the tools it exposes.

Create External MCP Server​

Opens a form to configure an External MCP Server in the project.

To configure an External MCP Server:

  1. Give the server a name
  2. Enter the URL of the Streamable HTTP MCP server
  3. Optionally add headers to be sent to the server, such as authentication headers. Header values can be entered directly or backed by project secrets (see the Manage Secrets project command)
  4. Optionally test the connection to the server before saving
  5. Click Create

External MCP Server Commands​

Edit External MCP Server​

Opens the External MCP Server allowing for changes to be made to its configuration. The connection can be tested from the editor before saving.

View External MCP Server JSON​

Opens the full External MCP Server JSON Twinit configuration in a new file.

Delete​

Deletes an External MCP Server configuration from the project.

Knowledgebase Commands​

Delete Knowledgebase​

Deletes a Knowledgebase from your project.

Prompt Templates Folder Commands​

The Prompt Templates folder lists reusable prompt templates in the project. Expand the folder to see existing templates.

A prompt template is a centrally stored instruction that agents and MCP clients can reuse. There are two kinds:

  • Fixed — static text with no arguments
  • Dynamic — text with {{placeholder}} arguments filled at runtime

Assign a template to an agent from Create Agent or Edit Agent. The AI Service loads the template by its user type and injects it as the agent's instruction when a team runs.

For platform behavior, see Prompt templates in the Twinit concepts docs.

Create Prompt Template​

Opens a form to create a Prompt Template.

To create a Prompt Template:

  1. Right-click the Prompt Templates folder and select Create Prompt Template
  2. Give the template a name (letters, digits, underscores, and hyphens)
  3. Enter a user type (letters, digits, and underscores). This is the identifier agents use when the template is assigned
  4. Optionally add a description
  5. Enter the template text. Use {{argumentName}} placeholders for values supplied at runtime
  6. Optionally add arguments that describe each placeholder, or use Add missing from template to create argument rows from placeholders in the template text
  7. Click Create

Prompt Template Commands​

Edit Prompt Template​

Opens the Prompt Template so you can change its name, user type, description, template text, and arguments. Changing the user type can break agents that already use this template.

View Prompt Template JSON​

Opens the full Prompt Template JSON Twinit configuration in a new file.

Delete​

Deletes a Prompt Template from the AI Service. This cannot be undone.

Teams Folder Commands​

Fixed-flow teams and supervisory teams both live under the Teams folder. Supervisory teams are labeled {name} (Supervisory), use a distinct organization icon, and expand to supervisors with nested sub-agents. If a supervisor has a fallback agent, that agent is listed last under the supervisor and labeled Failure agent.

For platform behavior, see Teams in the Twinit concepts docs.

Create Teams​

Creates a new fixed-flow team by selecting and ordering Twinit AI Agents.

To create a new Team:

  1. Right click on the Teams folder in the AI Service panel and select Create Team
  2. Give the team a name, an optional description, and select one or more Twinit AI Agents to add to the team
  3. Click Submit
  4. If you selected more than one Twinit AI Agent a new input will open allowing to define the flow order of the agents in the team
  5. Click Confirm Order

Create Supervisory Team​

Creates a team where one or more supervisor agents route work to specialists at runtime instead of following a fixed flow. You can add multiple supervisors; each has its own sub-agents. The lowest-order supervisor is the entry point.

To create a supervisory team:

  1. Right-click the Teams folder and select Create Supervisory Team
  2. Give the team a name and an optional description
  3. Add at least one supervisor card. For each supervisor, choose the supervisor agent, one or more sub-agents, and optionally a fallback agent used if routing cannot settle. The fallback does not have to be a sub-agent
  4. Use Up / Down to set supervisor order
  5. Click Create

You need at least two agents with a user type in the project (one supervisor and one specialist). An agent can appear at most once as a supervisor or sub-agent.

Teams Commands​

Chat with Team​

Opens a new chat interface, allowing you to start a new conversation with the selected team. Works for both fixed-flow and supervisory teams.

On Twinit 5.2 and later, the chat streams progress as the team works: an activity log lists process updates, which agent is running, any configured invocation message, and tool names (not tool payloads). The team's answer appears as tokens arrive. Use Stop to cancel an in-progress stream. If streaming is not available, the chat falls back to a single reply when the team finishes.

Delete Team​

Deletes a Team from the Twinit AI Service. This cannot be undone. Works for both fixed-flow and supervisory teams.

Edit Team​

Edits a fixed-flow team to manage its properties, add/remove agents, and update agent flow order.

Edit Supervisory Team​

Edits a supervisory team: name, description, supervisors, sub-agents, order, and optional fallback agents.

Tools Folder​

The Tools folder displays a read-only list of the tools available to assign to Agents.