Connect Your Own Agent
Introduced in FlowFuse 3.0
Connect your own AI agent to FlowFuse to manage your platform and build and edit flows in your Node-RED instances. You bring the agent and the model it runs on.
Connect your agent
Any MCP client with HTTP transport support can connect. That is the only requirement.
Pick your agent for the address and the steps that apply to it:
01Copy the FlowFuse connector URL
You will paste this into your agent in the next step.
https://app.flowfuse.com/mcp02Copilot Studio, Tools, Add a tool
Choose Model Context Protocol and paste the URL. Describe what it is for: the orchestrator reads that to decide when to call it.
03Sign in and choose what it reaches
Which teams the agent may act on, and whether it has editing rights or read access only.
The same three steps, written out:
- Add the FlowFuse MCP address in your agent's connector settings. See where to add it, per agent for yours.
On FlowFuse Cloud:https://app.flowfuse.com/mcp
Self-hosted, substitute your own platform address:https://<your-flowfuse.domain>/mcp - Sign in. FlowFuse uses OAuth, so your agent sends you to a FlowFuse login page. If your client asks for an OAuth client ID or secret, leave them blank; FlowFuse registers your client for you.
- Choose what the agent may do. Signing in takes you to a FlowFuse authorization page. There you pick read-only or full access, scope it to all your teams or specific teams, and set an expiration date for the grant.
Your agent is now connected. If your client does not support OAuth, use a token instead, see clients without a sign-in flow.
Note: This is separate from MCP server nodes, which build MCP servers inside your flows. This page is about operating FlowFuse itself, where FlowFuse is the server and your agent is the client.
What your agent can do
A connected agent works in two areas:
- Platform automation: manage your teams, applications, hosted and remote instances, snapshots and more.
- Flow building: read, build and edit the flows in your Node-RED instances.
What it can do in a given team or instance reflects the access you granted. See the FlowFuse MCP server page for the full list.
Flow work runs in a live Node-RED editor session. When you ask for it, the agent guides you to connect a session; a control in the platform header selects which of your open browser sessions it works in.
Where to add it, per agent
The common agents and where their settings live. Any other MCP-over-HTTP agent connects the same way.
Microsoft Copilot
In Copilot Studio, open your agent's Tools page, select Add a tool, then New tool, then Model Context Protocol. Give the server a name and a description (the orchestrator uses the description to decide when to call it), and enter the FlowFuse MCP address as the server URL.
To make FlowFuse available across a Microsoft 365 tenant rather than in a single agent, a tenant administrator registers it in the Microsoft 365 admin center. Once approved it appears in Copilot Studio for everyone.
Access through Copilot Studio runs over Power Platform connectors, so any Power Platform data policy your organisation has also governs it.
ChatGPT
Custom connectors live behind developer mode. Turn it on under Settings, then Apps & Connectors, then Advanced settings, then add FlowFuse by URL and sign in. Developer mode requires a paid plan.
Claude
Where custom connectors are available on your plan, add one and enter the FlowFuse MCP address.
On Team and Enterprise plans an owner adds the connector for the organisation first, then each person connects and signs in individually.
Coding agents
A coding agent can add the connector to itself. Ask it, rather than editing its configuration by hand:
Add the FlowFuse MCP tool at https://app.flowfuse.com/mcp. Then ask me to complete the sign-in in the browser that opens.
This works in Claude Code and Codex, and in any other agent that can change its own MCP configuration. It also survives those clients changing how a remote server is added, which they do at different times and in different ways.
If you would rather add it yourself, Claude Code takes:
claude mcp add --transport http flowfuse https://app.flowfuse.com/mcp
Other command-line and editor agents
Cursor, Visual Studio Code and Gemini CLI all connect to the same address. Where a client supports OAuth, sign in; otherwise use a token, see clients without a sign-in flow.
Local and self-hosted models
Use any HTTP-capable MCP client, such as LM Studio, LibreChat or Open WebUI, pointed at your own model, and add the FlowFuse address as a server. Ollama is a model runtime, not an agent, so it needs an MCP client in front of it.
Clients without a sign-in flow
If your client does not support OAuth, give it a token in its configuration file instead. Both routes reach the same FlowFuse with the same enforcement.
Create a Personal Access Token, scope it to the team you want the agent working in, and give the client the FlowFuse address with that token as a bearer token in an Authorization header.
The config format is the client's, not FlowFuse's. Two JSON shapes are common, one keyed on servers and one on mcpServers; clients also differ on where the file lives and whether they accept headers. Follow your client's reference: the servers shape and the mcpServers shape.
Approvals and audit
Each FlowFuse tool carries its recommended usage and permissions, so a connected agent knows what it is for before calling it. Most MCP clients then ask you to confirm before running a tool. That prompt is the client's, not FlowFuse's, so its look and whether you can turn it off vary. FlowFuse Expert's own approval cards do not apply here.
What FlowFuse enforces on every call is your grant: the teams, and read-only or full access. Each tool carries the access it needs, and FlowFuse rejects a call whose tool reaches past your grant, whether the grant came from signing in or from a token's scope.
Actions an agent takes appear in the audit log, attributed to your account and marked as having come from a connected agent.
If something is not working
A change was refused. The agent has read-only access. Re-connect it and grant full access.
The agent cannot reach a team. That team was not included when you signed in. Re-connect and include it.
The agent cannot see the instance you mean. Flow and editor work runs in a connected editor session. Ask your agent to list your sessions and connect to the right one.
Requirements
An instance the agent works in should run a current launcher or Device Agent and in-editor assistant. These update when an instance restarts, so a long-running instance may lag. If the agent cannot do something you expected in an instance, this is usually why.
On self-hosted, platform messaging runs over the MQTT broker, so the Team Broker must be available; what that needs from you depends on how your platform was installed, see MQTT Broker configuration. AI features also require an Enterprise licence with AI enabled.