How to use the FlowEngine MCP server with Claude Code

If you write code with Claude Code, you've probably noticed that the agent is fluent at reasoning about your codebase but blind to your hosted infrastructure. It can describe how to deploy an n8n workflow but not actually deploy it. It can tell you which instance is misbehaving in theory but can't ask a real instance for its real status.

FlowEngine ships a Model Context Protocol (MCP) server that closes that gap. Wire it up once in your Claude Code config and the agent gains direct access to your FlowEngine account: list instances, provision new ones, deploy workflows, check status, manage components. This guide walks through the setup, the most useful tools, and a few examples that make the value obvious.

What MCP gives you here

MCP is the protocol Claude Code uses to talk to external services as first-class tools. When you add the FlowEngine MCP server, the agent can call into your FlowEngine account the same way it calls into ls or git. It doesn't ask permission for every call, it just does the right thing in context.

For agency work, three concrete unlocks:

• Conversational provisioning. "Spin up a new n8n instance for client Acme on the agency tier" becomes a single sentence, not a portal click-through.
• Status without context-switching. "Is the Stripe sync workflow running on Acme's instance?" answers itself without leaving your editor.
• Workflow operations from a code review. While reviewing a code component, you can deploy it directly to the right client instance without opening the FlowEngine portal.

Step 1: Get your FlowEngine API key

In the FlowEngine portal, go to Settings → API Keys and generate a new key. Copy it once — like every responsible secret manager, FlowEngine doesn't show it twice. Stash it in your password manager, you'll paste it into the MCP config in a moment.

Step 2: Add the MCP server to Claude Code

Claude Code reads MCP servers from ~/.claude.json (or the per-project .claude.json). Add a new entry under mcpServers:

{
  "mcpServers": {
    "flowengine": {
      "url": "https://flowengine.cloud/mcp/<your-api-key>",
      "transport": "http"
    }
  }
}

Replace <your-api-key> with the key from step 1. The MCP server is HTTP-based, no local install required.

Restart Claude Code (or run /restart inside the session). The agent now has access to a set of flowengine_* tools. Run /mcp to confirm the server is connected.

Step 3: Run your first MCP tool

The simplest verification: ask Claude Code to list your instances.

You: list my flowengine instances

The agent calls flowengine_get_instances and returns the list. Names, URLs, statuses, service types. If you see your real instance list, the MCP wiring works.

The most useful tools

The full tool list is in the FlowEngine MCP docs, but these are the ones that earn their keep daily for agency work:

• flowengine_get_instances — list everything you host. Takes no args.
• flowengine_get_instance_status — live status for a specific instance. Useful for "is this client's automation up right now?"
• flowengine_provision_n8n — deploy n8n to an empty instance slot. Takes instanceId.
• flowengine_deploy_instance — push a code update or repo to an instance.
• flowengine_manage_instance — start / stop / restart / redeploy. Useful for "this client's instance is misbehaving, restart it."
• flowengine_create_instance_backup — manual backup before a risky change.
• flowengine_get_instance_logs — pull recent logs without leaving the editor.
• flowengine_create_workflow — push an n8n workflow JSON straight to a client instance.

The agent uses these naturally given the conversation context. You don't have to call them by name — you describe what you want and Claude Code figures out which tool to use.

Three example sessions

Provisioning during code review. You're reviewing a PR that adds a new automation for a new client. After merging:

You: provision a fresh n8n instance for the new client "Beta Industries"
     and import the workflow from this repo's workflows/beta-onboarding.json

The agent provisions, waits for active status, and pushes the workflow. You're back to code review while the deployment happens.

Triage from logs. You're debugging a client's failing automation:

You: what does Acme's "Stripe sync" workflow look like in the last hour?
     show me the failed executions and their errors

The agent calls status, pulls logs, and summarizes. You decide on the fix without ever opening the FlowEngine portal.

Bulk migration. You're moving five clients to a larger server type:

You: list all instances on the "starter" plan, then redeploy each onto
     "pro" and confirm they're back to active

The agent does it sequentially, reporting progress. You stay in the editor.

Try it

If you build automations professionally and use Claude Code, this is the kind of integration that pays for itself the first time you avoid a context-switch into the FlowEngine portal during a debugging session. Sign up for FlowEngine, generate an API key, and add the MCP server to your ~/.claude.json. The free tier is enough to test the connection with a personal instance before wiring it into your agency workflow.