leverly-automation

Leverly Automation - Automating Leverly Tasks via Rube MCP

Skill Overview

Using Composio’s Rube MCP service, dynamically discover and execute Leverly tools to enable Leverly task automation within an MCP client.

Use Cases

Use Leverly in Claude Desktop or other MCP clients: No code required—directly call Leverly tools via MCP to complete business operations.

Workflows that must adapt dynamically to tool-architecture changes: Do not rely on hard-coded tool definitions. Search the latest architecture before each run to prevent workflow failures when tools are updated.

Batch processing Leverly tasks: Supports executing multiple tool operations at once, suitable for scenarios that require orchestration of multi-step business processes.

Core Capabilities

Dynamic tool discovery: Use RUBE_SEARCH_TOOLS to search for currently available Leverly tools, retrieving the latest tool architecture, input parameters, and execution recommendations to ensure the workflow always uses the correct tool definitions.

Connection and state management: Use RUBE_MANAGE_CONNECTIONS to manage Leverly account connections—check connection status, complete the authorization flow, and ensure that only connections in an ACTIVE state can execute tools.

Multi-tool orchestration and execution: Use RUBE_MULTI_EXECUTE_TOOL to execute multiple Leverly actions within a single session, supporting session ID reuse and in-session memory passing—ideal for multi-step collaborative business workflows.

Frequently Asked Questions

What is Rube MCP? Do I need an API key?

Rube MCP is Composio’s provided MCP server endpoint. It exposes Composio toolkits (such as Leverly) to MCP clients. When using it, simply add https://rube.app/mcp as the MCP server address in your client configuration—no API key is required. After connecting, you will automatically receive tools like RUBE_SEARCH_TOOLS and RUBE_MANAGE_CONNECTIONS.

Why do I always need to search for tools before executing?

Composio’s tool architecture is updated from time to time. Tool slugs, parameter fields, and recommended execution plans may change. If these definitions are hard-coded, the workflow may fail after a tool update. By using RUBE_SEARCH_TOOLS to fetch the current architecture first, you ensure each execution uses the correct tool definitions—this is the core design principle of this skill.

What if the Leverly connection status is not ACTIVE?

After calling RUBE_MANAGE_CONNECTIONS, if the connection status is not ACTIVE, it will return an authorization link. Open that link in your browser to complete the Leverly account authorization flow. Then call RUBE_MANAGE_CONNECTIONS again to check the status until it shows ACTIVE, after which you can execute tool operations. Authorization only needs to be done once; afterward, the connection will remain in the ACTIVE state.

Can I batch-execute multiple Leverly tools?

Yes. When using RUBE_MULTI_EXECUTE_TOOL, pass multiple tool objects in the tools array. Each object should include the tool_slug obtained from the search results and arguments that match the architecture. All tools will run sequentially within the same session, suitable for scenarios requiring orchestration of multi-step business processes.

What’s the difference between RUBE_REMOTE_WORKBENCH and RUBE_MULTI_EXECUTE_TOOL?

RUBE_MULTI_EXECUTE_TOOL is for directly executing known tool actions, suitable for typical workflows. RUBE_REMOTE_WORKBENCH provides a remote Python environment where you can run the run_composio_tool() function. It supports more complex logic branching, looping, and custom scripts—ideal for scenarios requiring dynamic decisions or batch processing.

How can I get the full architecture of a specific tool?

If the tool information returned by RUBE_SEARCH_TOOLS includes a schemaRef field, you can pass that reference value to RUBE_GET_TOOL_SCHEMAS to retrieve the complete input/output schema definitions. This is useful for scenarios where you need to precisely construct complex parameters.