Workflow Discover¶

The Workflow Discover tool searches existing workflows by requirements and returns scored matches with gap analysis and reuse recommendations. This helps agents avoid creating duplicate workflows and instead reuse or fork existing ones.

Component type: workflow_discover

How It Works¶

The tool queries all active, non-deleted workflows on the platform (excluding the caller's own workflow), scores each one against the provided requirements, and returns the top matches sorted by score. The scoring considers trigger types, node types, tools, tags, and description similarity.

Each match includes a recommendation based on its score:

Score	Recommendation	Meaning
>= 0.95	`reuse`	The workflow already meets your requirements -- use it as-is
>= 0.50	`fork_and_patch`	Close enough to fork and modify via Workflow Create fork mode
< 0.50	`create_new`	No good match -- create a new workflow from scratch

Each match also includes a gap analysis listing what the existing workflow is missing compared to your requirements (missing triggers, node types, or tools).

Ports¶

Outputs¶

Port	Type	Description
`result`	`STRING`	JSON with `matches` (scored results), and `total_searched`

Output Format¶

{
  "success": true,
  "matches": [
    {
      "workflow_id": 12,
      "slug": "data-pipeline",
      "name": "Data Pipeline",
      "description": "Fetches and processes data hourly",
      "score": 0.82,
      "recommendation": "fork_and_patch",
      "capabilities": {
        "triggers": ["trigger_schedule"],
        "node_types": ["agent", "code"],
        "tools": ["http_request", "run_command"]
      },
      "gaps": {
        "triggers": [],
        "node_types": ["switch"],
        "tools": ["web_search"]
      },
      "success_rate": 0.95,
      "execution_count": 47
    }
  ],
  "total_searched": 15
}

Configuration¶

This tool has no configurable settings.

Usage¶

Connect this tool to an agent via the green diamond tool handle. It pairs naturally with Workflow Create:

flowchart LR
    WD[Workflow Discover] -.->|tool| Agent
    WC[Workflow Create] -.->|tool| Agent
    Model[AI Model] -.->|llm| Agent
    Trigger[Chat Trigger] --> Agent

Tool Signature¶

workflow_discover(requirements: str, limit: int = 5) -> str

Parameters:

Parameter	Type	Default	Description
`requirements`	string	(required)	JSON string with search criteria
`limit`	int	`5`	Maximum number of results

Requirements Format¶

The requirements parameter is a JSON string with optional keys:

{
  "triggers": ["trigger_chat", "trigger_telegram"],
  "node_types": ["agent", "code", "switch"],
  "tools": ["web_search", "run_command"],
  "tags": ["support", "automation"],
  "description": "A workflow that handles customer support queries",
  "model_capability": "gpt-4"
}

Key	Type	Description
`triggers`	array	Required trigger types
`node_types`	array	Required node component types
`tools`	array	Required tool component types
`tags`	array	Tags to match against
`description`	string	Natural language description for similarity matching
`model_capability`	string	Required model capability substring

All keys are optional. Omitted keys are not factored into the score.

Example¶

An agent tasked with building a new workflow first checks for existing ones:

User: Build me a workflow that monitors a website and sends Telegram alerts.

Agent thinking: Let me check if something similar already exists.

Tool call: workflow_discover(
    requirements='{"triggers": ["trigger_schedule"], "node_types": ["agent"], "tools": ["http_request"], "tags": ["monitoring"], "description": "Monitor a website and send alerts via Telegram"}'
)

Result: {
    "success": true,
    "matches": [
        {
            "slug": "site-monitor",
            "score": 0.72,
            "recommendation": "fork_and_patch",
            "gaps": {"tools": [], "node_types": [], "triggers": []}
        }
    ],
    "total_searched": 8
}

Agent: I found an existing workflow "site-monitor" with a 72% match.
       I'll fork it and add Telegram delivery.

The agent then uses workflow_create with based_on: "site-monitor" and patches to add Telegram notification.

Discover-Then-Create Pattern

The recommended pattern is: discover first, then either reuse (score >= 0.95), fork and patch (score >= 0.50), or create new (score < 0.50). This prevents workflow sprawl and promotes reuse of tested, proven workflows.

Scoring Includes Execution History

The scoring algorithm considers each workflow's historical success rate and execution count. Workflows with a high success rate and many successful executions score higher, as they are more likely to work reliably.