How to move from Copilot Studio to FlowRunner

A Copilot Studio agent leaves Copilot Studio in a solution file, or as YAML through the VS Code extension. Those files hold the agent’s instructions, its topics, and its agent flows. FlowRunner Prompt to Flow reads them in the AI tool you already use and builds the same agent as a FlowRunner flow: an AI Agent block with the same instructions, the tools it needs, and each agent flow as a flow the agent can call.

You move one agent at a time. What’s left for you is what no export carries: your model key, connected accounts, the knowledge documents, and testing before you switch.

If the agent calls Power Automate cloud flows that live outside its solution, move those first with the Power Automate guide, so the agent’s tools exist when you rebuild it.

1. Export your agents from Copilot Studio

Agents leave only inside a custom unmanaged solution. The default solution can’t be exported. You need at least the System Customizer security role.

1. Create a solution, choose Add existing > Agent > Agent, select the agents, and click Add.
2. On the agent’s three dots, choose Advanced > Add required objects. That pulls in its topics, flows, and environment variables.
3. In the solution explorer, select the solution and choose Export solution. Pick Managed or Unmanaged, and download the .zip.

The file holds the agent’s instructions, its topics, and the components you added: agent flows, environment variables, connection references, and custom connectors.

Some things aren’t in it:

• Connections and sign-in. Connections and user authentication have to be set up again.
• Some knowledge. Microsoft notes that custom topics and knowledge sources might not be included if they’re stored as separate resources or Dataverse tables. Keep the original documents.
• Comments, IDs, and channel details. Topic and node comments don’t travel, and the icon and channel details may be empty.

Two exports fail outright. A topic name with a period in it blocks the export, so rename it first. An agent with more than 250 topics or 100 entities needs the classic export experience.

There’s a second route. The Copilot Studio extension for VS Code clones the agent as YAML files: agent.mcs.yaml, one file per topic under topics/, and each agent flow as workflows/<name>/workflow.json. Uploaded knowledge files aren’t downloaded automatically; you pick them in the Remote Knowledge Files window. For one topic, More > Open code editor shows its YAML, which you can copy.

Before you leave, save a handful of real questions from the agent’s test chat or its conversations, with the answers it gave. You’ll test with them in step 5.

2. Connect FlowRunner Prompt to Flow to your AI tool

FlowRunner Prompt to Flow is an MCP server you add to the AI tool you already use. You connect it once and sign in with your FlowRunner account. The server URL is the same for every tool: https://app.flowrunner.ai/mcp.

• Claude (web and desktop): open Settings > Connectors, click + Add, then Add custom connector. Enter FlowRunner and the server URL, click Add, then Connect, and approve the FlowRunner sign-in. On a Team or Enterprise plan, an Owner adds it once under Organization settings > Connectors.
• Claude Code: run these two commands in your project folder, and approve the sign-in page that opens:

claude mcp add --transport http flowrunner https://app.flowrunner.ai/mcp
claude mcp login flowrunner

• ChatGPT (on the web): open ChatGPT Plugins, click Add, then Add custom MCP server. Enter FlowRunner, put the server URL under Connection, choose OAuth, and approve the sign-in. In a chat, type @ and pick FlowRunner.
• Codex: add the server to ~/.codex/config.toml, then run codex mcp login flowrunner:

[mcp_servers.flowrunner]
url = "https://app.flowrunner.ai/mcp"

• Cursor: add the server to ~/.cursor/mcp.json, then open Customize > MCPs and follow the sign-in prompt:

{
  "mcpServers": {
    "flowrunner": { "url": "https://app.flowrunner.ai/mcp" }
  }
}

If you don't have a FlowRunner account yet, the Free plan needs no card, and every new account starts with 14 days of Professional. The setup guide in the docs has each tool step by step.

3. Convert one agent at a time

Start a new chat, paste this request, then paste or attach the agent’s files after it. From a solution .zip, unzip it first. From the VS Code extension, paste agent.mcs.yaml, the topic files, and any workflow.json files.

Rebuild this Microsoft Copilot Studio agent as a FlowRunner flow. The files come from a Copilot Studio solution export (or the VS Code extension’s YAML files): the agent’s instructions, its topics with their trigger phrases and nodes, and any agent flows as workflow.json in Power Automate’s workflow definition format. Keep the agent’s instructions and the same behavior. Build each agent flow as its own FlowRunner flow the agent can call as a tool. Connections, sign-in settings, and knowledge files aren’t in the export. Before you build, tell me anything you can’t map one to one, and ask me about anything the files don’t say.

If the agent asks people for input or approval, add a second paragraph:

This agent uses [Request for information / a multistage approval]. Build that as a step that reaches [who] on [email / Slack / WhatsApp / phone] with the details they need, and waits for their answer before the agent carries on. Run the agent on my own [provider] key.

Then add a line or two about what the files can’t say: where the knowledge documents live, and how people reach the agent today (Teams, a website, a schedule), so the AI can pick how the flow starts.

Answer its questions, and it builds the flow in your workspace. Topics that only steer the conversation usually fold into the agent’s instructions; topics that send work down different paths become branches, often an AI Router that routes by meaning. One agent per chat keeps the questions specific.

4. Finish what the export can’t carry

Each conversion ends with a short list from the AI. Expect some of these:

• Add your model key. FlowRunner runs every agent on your own key. Add it once in the workspace and every flow can use it.
• Connect your accounts. Each app is connected once in OAuth Connections and shared across your flows.
• Attach the flow tools. The AI tells you which flows to attach to the agent as tools. Open the AI Agent block, go to Manage Capabilities > Flows, and add each one. The flows need to be live.
• Add forms. If an approval or an intake step should be a Form, the AI tells you, and you build it in the flow editor.
• Load the knowledge. Create a Knowledge Base, upload the agent’s documents, and attach it to the agent.
• Register MCP servers. If the agent used MCP servers, add each one under Agent tools & Knowledge > MCP Servers and attach its tools to the agent (how).
• Decide how the flow starts. In Copilot Studio, people reached the agent in Teams, Microsoft 365 Copilot, or a website. In FlowRunner the flow can start from an API call from your app, a form, an incoming email, a Slack message, or a schedule.
• Start the flow, if the AI left it stopped.

5. Test, then switch over

Run the flow with the questions you saved in step 1 and compare the answers. With FlowRunner Prompt to Flow connected, you can ask your AI tool to run it, for example “Run my Benefits helper flow with: How many vacation days do I have left?” Each block’s inputs and outputs for every run are in the flow’s run history. When the answers match, point people and apps at the FlowRunner flow, watch a few days of runs, and turn the Copilot Studio agent off. Keep the solution file.

How Copilot Studio maps to FlowRunner

Copilot Studio	FlowRunner	What changes
Agent and its instructions	AI Agent block with the same system prompt	The agent is a block inside a flow, and runs on your own model key
Topics	Agent instructions, or branches with an AI Router	Routing shows on the canvas instead of in trigger phrases
Agent flow (“When an agent calls the flow”)	A flow the agent calls as a tool (flows as agent tools)	Same execution whether an agent, a schedule, or an API starts it
Respond to the agent	Return Result	The agent waits for the flow and reads what it returns
Request for information, multistage approval	A step that asks by email, Slack, WhatsApp, phone, or a Form, then waits at an External Callback	The person can be anyone, in their own channel
Knowledge sources	Knowledge Base, or the SharePoint and OneDrive integrations as tools	You upload the documents again
Connectors and tools	Integration actions attached to the agent	2,100+ integrations, all on every plan
MCP servers	MCP Servers registered in the workspace	Attached to the agent as tools
Connection references	Connections, set once per workspace	Connect each account again
Channels (Teams, website)	API call, form, email, Slack message, or schedule	You choose how the flow starts
Copilot Credits	Executions, one per complete run	Model tokens go on your provider’s bill

Quick wins on day one

• One meter. Every run is one execution, however many tools the agent calls. Growth is $45 for 12,000 runs, $0.00375 a run, in place of separate credit rates for answers, actions, flow steps, and tokens.
• Your own model key. Agents run on Anthropic, OpenAI, Google, or another supported provider, with no markup.
• Approvals in more channels. Email, Slack, WhatsApp, phone, or a Form, to anyone, on every plan.
• Every integration on every plan, including Teams, Outlook, and SharePoint, with unlimited users and workflows.
• Start on Free. 100 executions a month, no card, any real email address, and 14 days of Professional on signup.