Mid-market B2C teams can use Claude or ChatGPT to build and run respond.io AI Agents and Workflows and manage contacts and conversations within their set permissions.
Your team can use Claude or ChatGPT to turn a description of how customer conversations should work into AI Agents and automations in respond.io. Respond.io is a customer conversation platform that connects to both assistants through its MCP server.
For example, a sales manager could ask: “Create a sales AI Agent, attach our product information and let it update contact fields and assign conversations to Sales.” The assistant can create the agent, attach existing Knowledge Sources, the business information it answers from, and configure the actions it can take during a conversation. The team can also use the assistant to improve its instructions, activate it or change its setup as sales needs evolve.
The same approach works for Workflows, the automations that follow rules your team sets: “Create a Workflow that adds a tag to new contacts and assigns them to the Sales team.” The assistant builds a draft from the request or a template, then can validate and test it before the team approves publication. Managers can describe the process they need and review a working draft instead of configuring each step themselves.
Scheduled and time-based Workflow triggers are not supported through MCP. Google Sheets steps and file attachments still need finishing in the Workflow Builder.
Claude or ChatGPT can also carry out the daily work that surrounds those automations. A manager can ask it to find contacts matching specific criteria, review a lead’s message history and delivery status, update their record and send a follow-up with a template or attachment.
When a lead needs another team, the assistant can assign the conversation and add an internal comment explaining the next step. It can also open or close conversations, manage contact records and tags, and update Lifecycle stages, which track leads, customers and returning customers through their journey. This connects reviewing a customer’s situation with taking the next action.
The assistant can view teams, members and messaging apps and create custom fields and tags, so the team can adapt the information used by its agents and automations from the same place.
The MCP connection uses OAuth and follows the connected user’s existing permissions. Each connection covers one workspace, so an assistant acts within that user’s access rather than gaining control of every workspace.
Workflow validation flags configuration errors, and testing lets the team check behavior before launch. Tests can send real messages. Publishing, stopping or deleting a Workflow each requires confirmation. Editing a published Workflow stops it first and removes contacts currently in it; republishing does not re-enroll them. These controls let managers review the automation and approve changes that affect live conversations.
Connecting Claude or ChatGPT gives your team a way to operate respond.io through requests. It does not turn your existing assistant into an agent that automatically answers incoming customer messages.
Respond.io’s AI Agents run the customer conversations across messaging apps: replying, qualifying inquiries, recommending products and booking appointments, with human handoff when needed. Claude or ChatGPT can build and manage those agents and their Workflows, while respond.io runs them.
Supported MCP actions do not consume respond.io AI Credits. AI Agents still consume credits when replying to customers, including agents created or updated through MCP.