MCP: To create and manage Workflows
N
Nicholas Chuah
Business Problem
Businesses often know the automation outcome they want but still need to manually configure every Workflow through the UI. Creating triggers, connecting steps, setting up branches, validating the graph, testing different paths, and publishing safely can be time-consuming and requires detailed knowledge of the Workflow builder.
Without agent-assisted Workflow creation and validation, users may create incomplete configurations, miss required connections, introduce unintended side effects during testing, or publish changes without a clear understanding of their impact. Managing existing Draft or Stopped Workflows also requires users to inspect and troubleshoot them manually.
Desired Outcome
Enable an AI agent to turn an approved automation goal into a working Workflow through a controlled, permission-aware process.
The agent should be able to:
- Propose and create a Draft Workflow using supported Triggers, Steps, Branches, and connections.
- Inspect and edit Draft or Stopped Workflows while respecting user permissions and protecting existing versions.
- Validate the complete Workflow graph and return actionable errors, warnings, and publish blockers.
- Test representative paths using the supported test mechanism, with side effects isolated where the platform allows.
- Present a clear publish summary and require explicit user confirmation before publishing.
- Publish an attributable Workflow version, inspect authoritative run status, and support eligible retries or recovery to a previous version where the platform supports it.
Use Cases
- An onboarding specialist describes a business goal, such as qualifying new leads and assigning them to the right team, and the agent creates a Draft Workflow for review.
- A user asks the agent to inspect a Stopped Workflow, identify the failure, and propose or apply a permitted fix without overwriting the existing version.
- A user validates a Workflow before publishing and receives clear information about missing connections, invalid configurations, warnings, and blockers.
- A user tests representative branches without triggering unintended customer messages or other side effects where isolation is supported.
- A team reviews the proposed changes, affected paths, and expected behavior before explicitly approving publication.
- After publishing, users can identify which agent or user created the version, inspect the authoritative run status, retry eligible runs, or recover a previous version where supported.
K
Kantu Dev Smartbot
In our case, we need to build and manage workflows using MCP to automate customer service processes and WhatsApp campaigns using natural language instructions. It would be especially helpful for us to be able to create, edit, duplicate, validate, activate, and deactivate workflows from tools such as ChatGPT, Claude, or n8n, while adhering to Workspace permissions and requesting confirmation before publishing changes.
This functionality would significantly reduce setup time and make it easier to maintain complex automations without relying exclusively on the visual editor.
N
Nabilah Binti Salleh
Merged in a post:
AI: Create a workflow with AI
V
Vovwe Enyoyi
Desired outcome:
Customers would be able to create a Workflow with written instructions for example, to create a welcome workflow that intelligently routes users to the support and sales teams based on their response to the question, "How can I help you today? allowing you to automate and streamline your workflow creation process.
N
Nabilah Binti Salleh
Merged in a post:
MCP: To create and manage Workflows
Digi Arabia Marketing Team
Business Problem:
Teams managing large, complex automation setups — such as multi-step patient communication flows integrated with Hospital Information Systems — face a significant bottleneck: every workflow must be created and edited manually through the visual UI builder. Even minor changes to existing flows require going step-by-step through the builder, making iteration slow and error-prone at scale. There is currently no way to manage workflows programmatically, whether via API, MCP Server, or AI-assisted tooling.
Desired Outcome:
- Expose workflow management operations via a respond.io MCP Server and/or API, covering the full CRUD surface that currently exists only in the UI - Specifically, the requested MCP tools include list_workflows, create_workflow, update_workflow_step, and clone_workflow.
- AI-assisted workflow implementation — a native integration path (e.g., via Claude) that can translate natural-language instructions into valid workflow configurations, accelerating setup for teams building at scale.
Workaround:
Workflows can currently be exported as a JSON file and re-imported to duplicate or transfer them across workspaces. In theory, the exported JSON could be edited externally and re-imported to simulate a programmatic update. However, this is impractical for external systems or AI to generate valid JSON reliably without reverse-engineering. There is no API endpoint to import a workflow programmatically, so a human must still complete the final step manually.
A
Alex Monckton
This would be extremely useful for tailoring the conversation to specific business needs.