Tools & Integrations

Tools give Lyro real-world abilities - looking up an order, creating a ticket, checking delivery status, or calling your own API. This article covers the four ways you add tools and how they reach your agents.

The Tools hub

Open Tools in your dashboard to manage everything in one place. The page is organized into tabs:

• Integrations - pre-built connectors to popular third-party services.
• MCP Servers - remote Model Context Protocol servers you connect by URL.
• Custom Tools - your own HTTP endpoints, defined step by step.
• Secrets - encrypted credentials your custom tools can reuse.

A "Personalised" tab also appears once you have personalized tools configured.

Pre-built integrations

The Integrations tab lists ready-made connectors grouped by category - Payments, Communication, CRM, Analytics, Productivity, and Development. You can search by name and filter by category.

To connect one:

1. Click an integration to open its detail panel.
2. Click Connect. Depending on the service, you either authorize through a popup or enter credentials (such as an API key) directly.
3. Once connected, Lyro fetches the available actions the integration exposes.

Each action shows a risk badge so you know what it does:

You can disconnect an integration at any time from the same panel.

MCP (Model Context Protocol) servers

MCP lets you connect a remote server that exposes its own set of tools. Lyro speaks the Streamable HTTP variant of MCP.

To add one, open MCP Servers, click Add Server, and provide:

• Name - a label such as Linear, GitHub, or Internal API.
• Server URL - the server's HTTP endpoint, for example https://mcp.example.com/mcp.
• Description (optional) - a note about what the server provides.
• Authentication - how Lyro authenticates to the server.

Authentication options

Discovering tools

After adding a server, open it and click Discover Tools. Lyro connects using your auth settings and fetches the list of tools the server offers, showing each tool's name and description. A health badge (Healthy, Unhealthy, or Unknown) tells you whether the last connection succeeded.

Re-run Discover Tools whenever the remote server adds or changes its tools so Lyro stays in sync.

For security, server URLs must be public http(s) addresses. Internal or private network addresses are rejected.

Custom HTTP tools

When no integration or MCP server fits, build your own. Open Custom Tools and click New Custom Tool to open the editor.

A custom tool is made of:

• Name and description - keep the description short; use a Skill to explain when the agent should call the tool.
• Parameters - the inputs the agent fills in. Each has a snake_case name, a type (string, number, integer, or boolean), a required flag, and a description.
• Steps - one or more HTTP requests. Each step has a method (GET or POST), a URL, headers, and an optional JSON body.
• Response - what the tool returns: either the last step's output, or a custom template you write.

Referencing values in your tool

Anywhere in a URL, header, or body you can insert dynamic values with {{ }} templates:

• {{params.order_id}} - a parameter the agent supplied.
• {{secrets.my_token}} - a workspace secret (see below).
• Output from a prior step, so you can chain requests together.

You can test a tool right in the editor by entering sample parameter values and running a step, then save its output as a sample to reference downstream. Tools start as a draft and become ready once they pass validation.

Workspace secrets

Secrets are reusable, encrypted values - API keys, tokens, partner credentials - that your custom tools inject without exposing them in plain text.

In the Secrets tab, click Add secret and provide a snake_case name, the value, and an optional description. Reference it from any custom tool field as {{secrets.name}}.

Key behaviors:

• Values are encrypted at rest and write-once - they are never shown again after creation.
• To rotate a secret, delete it and re-create it with the same name.
• Deleting a secret that a tool still references will cause that tool to fail until you replace it. Lyro warns you which tools use it.

How tools reach your agents

Adding a tool does not make Lyro use it automatically. Tools are assigned through Skills, and Skills run on agents:

1. Create or open a Skill on the Agents page.
2. Add the tool (an integration action, MCP tool, or custom tool) to that Skill.
3. Make sure the Skill is active and attached to an active agent.

A custom tool that is marked ready but not in any active Skill on an active agent will never run - the editor shows a warning when this happens, so the path from tool to agent is always clear.

Use the Skill's instructions to tell the agent exactly when to reach for each tool. Then try it in the Playground before going live.

Next steps

• Build your agents and Skills
• Train Lyro on your content
• Developer APIs and webhooks