Flowcastle MCP server
{
"mcpServers": {
"flowcastle": {
"type": "http",
"url": "https://api.flowcastle.ai/api/mcp"
}
}
}{
"mcpServers": {
"flowcastle": {
"type": "http",
"url": "https://api.flowcastle.ai/api/mcp"
}
}
}{
"mcpServers": {
"flowcastle": {
"type": "http",
"url": "https://api.flowcastle.ai/api/mcp"
}
}
}{
"servers": {
"flowcastle": {
"type": "http",
"url": "https://api.flowcastle.ai/api/mcp"
}
}
}{
"mcpServers": {
"flowcastle": {
"type": "http",
"url": "https://api.flowcastle.ai/api/mcp"
}
}
}file .mcp.json (in your project root)file claude_desktop_config.jsonfile ~/.cursor/mcp.jsonfile .vscode/mcp.jsonfile ~/.codeium/windsurf/mcp_config.jsonversion 1.0.0
What it does
An MCP server for building, editing, and deploying Telegram bots on FlowCastle's visual flow platform. Developers can describe bot behavior in natural language and have an AI agent assemble the flow through MCP tools, or reference 23 production bot templates as JSON examples.
- Ask your AI coding agent to read your workspace and add a new flow—like a lead-qualification bot that greets users, asks questions, and saves responses—without writing bot code.
- Browse and apply pre-built bot templates (lead bots, support bots, digest bots) from the gallery, either one-click in the UI or programmatically via MCP search and apply tools.
- Study real flow structures to understand how to wire triggers, conditions, AI steps, payments, and integrations in your own bots.
Tools this server exposes
| Tool | Does |
|---|---|
| get_workspace_summary | Retrieve a summary of your FlowCastle workspace. |
| get_flow_context | Retrieve context and details of a specific flow. |
| validate_actions | Check a batch of flow-builder actions for correctness before applying. |
| apply_actions | Apply a batch of flow-builder actions (create flow, create block, create link, create variable) to your workspace. |
| search_flow_examples | Search for available bot templates. |
| get_flow_example | Retrieve a complete bot template as a batch of actions. |
Before you enable it
An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.