Falcon Builder MCP Server
Build, edit, and publish production AI agents from Claude, Cursor, ChatGPT, VS Code, or any MCP client. The Falcon Builder MCP server lets your assistant inspect executions, edit and duplicate workflows, run them, and publish them in your Falcon Builder workspace.
Available on every plan, including Free.
- Endpoint:
https://www.falconbuilder.dev/api/mcp(Streamable HTTP) - Docs: https://www.falconbuilder.dev/docs/mcp
- Registry name:
dev.falconbuilder/falcon-builder
This is a hosted server. Nothing to install or run locally.
Connect
OAuth clients: claude.ai, Claude Desktop, ChatGPT
No token needed. The client registers itself automatically.
claude.ai / Claude Desktop
- Settings → Connectors → Add custom connector
- Server URL:
https://www.falconbuilder.dev/api/mcp - Leave OAuth Client ID and Client Secret empty
- Pick a workspace and the permissions to grant, then approve
ChatGPT Settings → Connectors → Create (turn on Developer Mode if prompted). Same URL, OAuth fields left empty.
Token clients: Claude Code, Cursor, VS Code, Gemini CLI
Create a token first: Falcon Builder → Settings → Developer, name it, choose permissions, and copy it. Tokens are shown once. Each workspace can have up to 10 active tokens.
Claude Code
claude mcp add --transport http falcon-builder \
https://www.falconbuilder.dev/api/mcp \
--header "Authorization: Bearer YOUR_TOKEN"
Cursor (.cursor/mcp.json)
{
"mcpServers": {
"falcon-builder": {
"url": "https://www.falconbuilder.dev/api/mcp",
"headers": { "Authorization": "Bearer YOUR_TOKEN" }
}
}
}
VS Code (.vscode/mcp.json). VS Code prompts for the token and stores it securely.
{
"inputs": [
{
"type": "promptString",
"id": "falcon-token",
"description": "Falcon Builder API token (Settings > Developer)",
"password": true
}
],
"servers": {
"falcon-builder": {
"type": "http",
"url": "https://www.falconbuilder.dev/api/mcp",
"headers": { "Authorization": "Bearer ${input:falcon-token}" }
}
}
}
Gemini CLI (~/.gemini/settings.json). Use httpUrl; in Gemini CLI, url means SSE.
{
"mcpServers": {
"falcon-builder": {
"httpUrl": "https://www.falconbuilder.dev/api/mcp",
"headers": { "Authorization": "Bearer YOUR_TOKEN" }
}
}
}
Permissions
You choose these when you connect over OAuth or create a token. Read is always included.
| Permission | What it allows |
|---|---|
| Read | Inspect workflows, executions, logs, version history, and interfaces |
| Edit drafts | Change workflow drafts and configure interfaces without touching production |
| Run workflows | Execute workflows with test input. Runs count against your plan limits |
| Publish & rollback | Deploy drafts to production and restore earlier versions |
Tools
Read
whoami, switch_workspace, list_agents, list_workflows, get_workflow, get_workflow_definition, get_workflow_node, validate_workflow, list_executions, get_execution, get_execution_logs, list_workflow_versions, list_interfaces, get_interface
Edit drafts
create_workflow, edit_workflow, duplicate_workflow, duplicate_agent, create_interface, update_interface, add_interface_access_rule, remove_interface_access_rule, delete_interface
Run workflows
run_workflow, rerun_executions
Publish & rollback
publish_workflow, rollback_workflow, publish_interface
Every tool declares MCP annotations (read-only, destructive, idempotent, open-world), so clients that honor them can ask before anything changes. run_workflow and rollback_workflow also require an explicit confirm: true.
Things to ask your assistant
- "Why did the last run of my intake workflow fail? Show me the node that errored."
- "Duplicate my support agent and change its escalation step to email the on-call address."
- "Validate the draft of my lead-routing workflow, then publish it."
- "Roll my onboarding workflow back to the previous version."
About
Falcon Builder is built by NeoSky AI.