MCP Integration (ChatGPT, Claude, Gemini)

Connect Autoflowly to ChatGPT, Claude, Gemini, or any Model Context Protocol client — build apps, staff AI teams, run loops, and branch apps, all from a conversation in your AI assistant of choice.

There's nothing to install and no server to run — connect to the hosted server at https://mcp.autoflowly.com.

Connect Claude, Gemini, or any generic MCP client

Add a remote MCP connector pointed at:

OAuth is discovered automatically — you'll be prompted to sign in with your Autoflowly account on first use. No manual client ID/secret entry required.

Connect ChatGPT (Apps SDK)

In ChatGPT Developer Mode → Create App (not "Add MCP Server"), enter the manifest URL:

This gives you the richer experience — inline interactive widgets (live app preview, apps list, loops, branches) rendered directly in the ChatGPT conversation.

Try It Without Your Own Account

What You Can Do

Category	Tools	Example prompt
Build	conversational_mvp_chat, create_mvp_from_prd, quick_mvp	"Build me a task management app for remote teams"
App Dashboard	get_mvp_details, get_mvp_urls, redeploy_mvp, upgrade_mvp	"What's the status of my app?"
AI Team & Loops	list_apps, staff_app, start_loop, list_loops	"Hire an AI team for my shop, then run a loop to get 10 sales"
App Branching	list_branches, create_branch, activate_branch, delete_branch	"Create a branch called redesign and open its preview"
Custom Agents	create_agent, chat_with_agent	"Create a support agent and let me talk to it"
Templates & Analysis	list_templates, recommend_template, validate_startup_idea	"What template fits a subscription box business?"

The live server is the source of truth for the exact tool list at any moment — ask your AI assistant to run tools/list if you want the full current set.

How It Works

The MCP server is a thin, faithful wrapper over the same backend the web app uses — every tool call maps directly to a real API endpoint (see API Reference). It doesn't have its own separate feature set that can drift from what's actually deployed.

Self-Hosting / Development

Only relevant if you're developing the MCP server itself, not for normal use — see the mcp-server README in the repo.