Skip to main content
Not sure whether to use tools or sub-agents? Read Extending Airia Agent first to pick the right approach.

Add a tool

1

Open the Tools tab

Navigate to Settings → Airia Agent → Tools.
2

Browse available tools

The tool picker shows all tool definitions available in your tenant. Use the search box to filter by name or category.
3

Select tools

Click the tools you want to add. Selected tools appear in the Configured Tools list.
4

Bind credentials (if required)

Some tools require authentication credentials (for example, a CRM integration that needs an API key). If prompted, select an existing credential from the dropdown or ask an admin to add one via Credential Management.
5

Save

Click Save. The tools are available to Airia Agent immediately.

Enable or disable individual tools

After adding tools, you can turn individual tools on or off without removing them from the configuration. This is useful during rollout or troubleshooting.
  1. In the Tools tab, find the tool in the Configured Tools list
  2. Use the toggle next to the tool name to enable or disable it
  3. Click Save
Tool availability changes take effect on the next user message. Active conversations are not interrupted.

Connect an MCP server

MCP (Model Context Protocol) servers expose multiple tools through a single connection. Connecting one makes all of its tools available to Airia Agent.
1

Open the Tools tab

Navigate to Settings → Airia Agent → Tools.
2

Add an MCP server

Click Add MCP Server. Enter the server’s connection details:
  • Name: A display name shown in the tool list
  • URL: The MCP server endpoint
  • Authentication: Select the credential to use for server authentication, if required
3

Review available tools

Once connected, the platform fetches the server’s tool manifest. Review the tools it exposes and enable or disable individual ones as needed.
4

Save

Click Save.

Tool response history

The Save tool responses to history toggle controls whether the full output of each tool call is included in subsequent conversation turns for LLM context.
SettingBehavior
Off (default)Tool results are used to generate the immediate reply but are not stored in the conversation history passed to the model on future turns
OnFull tool call inputs and outputs are appended to conversation history, giving the model richer context across turns
When to enable: Turn this on for multi-step workflows where later tool calls benefit from knowing the output of earlier ones — for example, an agent that queries a database, processes the results, and then queries again based on what it found. When to leave off: Single-step tool calls and high-volume use cases where the added history tokens would increase cost or latency without meaningful benefit.

MCP placeholder bindings

When upgrading from an earlier version of Airia Agent to v2 or later, the Tools tab may display a Placeholder Bindings section requiring resolution before the upgrade can complete. Placeholder bindings are named slots in the new agent pipeline that must be mapped to actual MCP servers or tools in your configuration. Think of them as required connections the new version expects but that the upgrade process cannot automatically infer from your existing setup.
1

Open the Tools tab after triggering an upgrade

If placeholder bindings exist, a warning banner appears at the top of the Tools tab.
2

Review each binding

The list shows each placeholder name and its expected type (MCP server or tool). Match each one to the correct resource from your configured tools and servers.
3

Save bindings

Click Save. Once all bindings are resolved, the upgrade can proceed.
See Version Management for the full upgrade flow.

Best practices

  • Start small. Add two or three high-value tools first and validate that Airia Agent uses them correctly before adding more. A large tool list can increase latency and make it harder to debug unexpected behavior.
  • Use descriptive tool names. Airia Agent selects tools based on their names and descriptions. Tools with clear, specific descriptions are used more accurately.
  • Credential hygiene. Use service-specific credentials rather than shared personal credentials. This makes auditing easier and limits blast radius if a credential is rotated.
  • Monitor usage. Use Feeds to review which tools are being invoked and whether they are returning errors.