home / skills / composiohq / awesome-claude-skills / tavily-automation

tavily-automation skill

safe

/tavily-automation

This skill automates Tavily tasks via Rube MCP, ensuring tool schemas are fetched first to optimize workflows.

npx playbooks add skill composiohq/awesome-claude-skills --skill tavily-automation

Review the files below or copy the command above to add this skill to your agents.

Files (1)

SKILL.md

2.8 KB

---
name: tavily-automation
description: "Automate Tavily tasks via Rube MCP (Composio). Always search tools first for current schemas."
requires:
  mcp: [rube]
---

# Tavily Automation via Rube MCP

Automate Tavily operations through Composio's Tavily toolkit via Rube MCP.

**Toolkit docs**: [composio.dev/toolkits/tavily](https://composio.dev/toolkits/tavily)

## Prerequisites

- Rube MCP must be connected (RUBE_SEARCH_TOOLS available)
- Active Tavily connection via `RUBE_MANAGE_CONNECTIONS` with toolkit `tavily`
- Always call `RUBE_SEARCH_TOOLS` first to get current tool schemas

## Setup

**Get Rube MCP**: Add `https://rube.app/mcp` as an MCP server in your client configuration. No API keys needed — just add the endpoint and it works.

1. Verify Rube MCP is available by confirming `RUBE_SEARCH_TOOLS` responds
2. Call `RUBE_MANAGE_CONNECTIONS` with toolkit `tavily`
3. If connection is not ACTIVE, follow the returned auth link to complete setup
4. Confirm connection status shows ACTIVE before running any workflows

## Tool Discovery

Always discover available tools before executing workflows:

```
RUBE_SEARCH_TOOLS
queries: [{use_case: "Tavily operations", known_fields: ""}]
session: {generate_id: true}
```

This returns available tool slugs, input schemas, recommended execution plans, and known pitfalls.

## Core Workflow Pattern

### Step 1: Discover Available Tools

```
RUBE_SEARCH_TOOLS
queries: [{use_case: "your specific Tavily task"}]
session: {id: "existing_session_id"}
```

### Step 2: Check Connection

```
RUBE_MANAGE_CONNECTIONS
toolkits: ["tavily"]
session_id: "your_session_id"
```

### Step 3: Execute Tools

```
RUBE_MULTI_EXECUTE_TOOL
tools: [{
  tool_slug: "TOOL_SLUG_FROM_SEARCH",
  arguments: {/* schema-compliant args from search results */}
}]
memory: {}
session_id: "your_session_id"
```

## Known Pitfalls

- **Always search first**: Tool schemas change. Never hardcode tool slugs or arguments without calling `RUBE_SEARCH_TOOLS`
- **Check connection**: Verify `RUBE_MANAGE_CONNECTIONS` shows ACTIVE status before executing tools
- **Schema compliance**: Use exact field names and types from the search results
- **Memory parameter**: Always include `memory` in `RUBE_MULTI_EXECUTE_TOOL` calls, even if empty (`{}`)
- **Session reuse**: Reuse session IDs within a workflow. Generate new ones for new workflows
- **Pagination**: Check responses for pagination tokens and continue fetching until complete

## Quick Reference

| Operation | Approach |
|-----------|----------|
| Find tools | `RUBE_SEARCH_TOOLS` with Tavily-specific use case |
| Connect | `RUBE_MANAGE_CONNECTIONS` with toolkit `tavily` |
| Execute | `RUBE_MULTI_EXECUTE_TOOL` with discovered tool slugs |
| Bulk ops | `RUBE_REMOTE_WORKBENCH` with `run_composio_tool()` |
| Full schema | `RUBE_GET_TOOL_SCHEMAS` for tools with `schemaRef` |

---
*Powered by [Composio](https://composio.dev)*

Overview

This skill automates Tavily tasks through Composio's Tavily toolkit using the Rube MCP orchestration layer. It provides a clear workflow for discovering current tool schemas, managing Tavily connections, and executing multi-tool operations reliably. The skill emphasizes always searching tools first and validating connection status before running workflows.

How this skill works

The skill instructs agents to call RUBE_SEARCH_TOOLS first to retrieve available tool slugs, input schemas, and recommended execution plans. It then verifies the Tavily connection via RUBE_MANAGE_CONNECTIONS and runs operations with RUBE_MULTI_EXECUTE_TOOL (including an explicit memory object and session ID). It includes guidance for pagination, schema compliance, and session reuse across a workflow.

When to use it

When you need to run Tavily operations programmatically through Composio and Rube MCP.
When tool schemas may have changed and you must discover the latest inputs before calling a tool.
When coordinating multi-step Tavily workflows that require connection validation and session state.
When executing bulk or parallel Tavily jobs via RUBE_REMOTE_WORKBENCH or multi-execute calls.
When building agents that must be robust to schema changes and connection states.

Best practices

Always call RUBE_SEARCH_TOOLS at the start of each workflow to fetch current tool slugs and schemas.
Verify the Tavily connection is ACTIVE with RUBE_MANAGE_CONNECTIONS before executing any tools.
Send a memory object (even if empty) in RUBE_MULTI_EXECUTE_TOOL calls to satisfy the MCP contract.
Reuse a session_id for related operations; generate a new session for distinct workflows.
Respect exact field names and types from search results—do not hardcode arguments or rely on outdated schemas.
Handle pagination tokens in search responses and iterate until all results are retrieved.

Example use cases

Discover and run a Tavily data transformation tool: search tools, validate connection, then execute with schema-compliant args.
Automate bulk Tavily tasks using RUBE_REMOTE_WORKBENCH and run_composio_tool() for parallel processing.
Create an agent that orchestrates a multi-step Tavily workflow, reusing a session_id and preserving memory across steps.
Fetch full tool schemas with RUBE_GET_TOOL_SCHEMAS when building integration tests or developer tooling.
Recover gracefully from connection issues by guiding users through the auth link returned by RUBE_MANAGE_CONNECTIONS.

FAQ

Do I need API keys to use Rube MCP?

No API keys are required; add https://rube.app/mcp as an MCP server endpoint in your client configuration.

What if a tool schema changes mid-workflow?

Always call RUBE_SEARCH_TOOLS before each new workflow or critical step to get the latest schema; validate inputs against returned schemas.

Can I omit the memory parameter in multi-execute calls?

No. Include a memory object ({} if empty) in RUBE_MULTI_EXECUTE_TOOL calls to comply with MCP expectations.