Use Wireflow from Claude Code, Cursor, Codex and ChatGPT

To use Wireflow from an AI agent, add the Wireflow MCP server, https://www.wireflow.ai/api/mcp, to your agent as a remote server and sign in with your Wireflow account. There is no API key to copy. Your agent can then build workflows, run them on your credits, read the results, and generate and edit images, video and audio across 100+ models.

This page has copy-paste setup for Claude Code, Cursor, Codex and ChatGPT, then a first prompt to test the connection.

Connection details

Setting	Value
Server URL	https://www.wireflow.ai/api/mcp
Transport	Streamable HTTP
Sign-in	OAuth 2.1 authorization code flow with PKCE. Clients register themselves (Dynamic Client Registration), so there is no client ID or API key to paste
Consent	You approve a consent screen that lists exactly the permissions the client asked for
Standard permissions	workflows:read, workflows:create, workflows:execute (spends credits), executions:read, models:read, usage:read, automations:read. A client that asks for nothing in particular gets the read and run scopes
Opt-in permissions	workflows:write, automations:write, teams:invite, social:publish, social:read. Withheld unless the client names them when it connects
Revoke	Any time, from your Wireflow settings

What your agent can do

Job	Tools
Check the connection	whoami, get_credit_balance
Find a starting point	list_workflows, get_workflow, list_templates, search_recipes, list_models
Build	create_workflow_from_prompt, clone_template, update_workflow_config
Run and read results	run_workflow, run_app, get_execution
Edit video	get_video_timeline, edit_video_timeline, preview_video_frames

create_workflow_from_prompt and clone_template save private drafts and run nothing. run_workflow and run_app spend credits on the connected account. preview_video_frames shows still frames of a video node without paying for a render. When a step has no Wireflow node, the agent can report the gap with send_feedback, which needs no scope and spends nothing. Every tool is listed on the MCP connector page.

Claude Code

Option 1: the Wireflow skills plugin. It adds the Wireflow connector and nine ready-made skills, such as AI image generation and AI video generation. In Claude Code, run:

/plugin marketplace add wireflowINC/skills
/plugin install wireflow-skills@wireflow

Then run /mcp, choose wireflow and sign in with your Wireflow account.

Option 2: the connector only. In your terminal, run:

claude mcp add --transport http wireflow https://www.wireflow.ai/api/mcp

Then run /mcp in Claude Code to sign in.

Claude.ai and Claude Desktop: add https://www.wireflow.ai/api/mcp as a custom connector.

Cursor

Add the server to ~/.cursor/mcp.json to use it in every project, or to .cursor/mcp.json inside one project:

{
  "mcpServers": {
    "wireflow": {
      "url": "https://www.wireflow.ai/api/mcp"
    }
  }
}

Cursor supports OAuth for servers that require it, so you sign in with your Wireflow account instead of adding a key or headers. Turn the server on or off from Customize in the Cursor sidebar. If it does not connect, open the Output panel and choose MCP Logs.

Codex

Add the server from your terminal, then sign in:

codex mcp add wireflow --url https://www.wireflow.ai/api/mcp
codex mcp login wireflow

Or add it to ~/.codex/config.toml yourself, then run codex mcp login wireflow:

[mcp_servers.wireflow]
url = "https://www.wireflow.ai/api/mcp"

The Codex CLI, the Codex IDE extension and the ChatGPT desktop app share this configuration. Check the server with codex mcp list, or type /mcp inside Codex.

ChatGPT

Add Wireflow from ChatGPT on the web:

1. Go to ChatGPT Plugins.
2. Select the plus button, then Add custom MCP server.
3. Enter the name Wireflow and, if you like, a description.
4. Under Connection, enter the server URL https://www.wireflow.ai/api/mcp.
5. For authentication, choose OAuth.
6. Review the risk warning, select I understand and want to continue, then select Create as a plugin.
7. Install the plugin before you use it in a conversation. When ChatGPT opens the Wireflow sign-in, sign in with your Wireflow account and approve the consent screen.
8. In the prompt box, type @ and select Wireflow.

ChatGPT asks you to confirm write actions by default. In a ChatGPT workspace, the workspace's permissions and security settings apply to adding and using custom MCP servers.

Add the Editor connector (optional)

Wireflow has a second MCP server for editing and rendering what is already in your account:

https://www.wireflow.ai/api/mcp/editor

It reads, edits and renders your workflows, video timelines and image composites. It does not generate images, video or audio. Rendering is deterministic and costs 0 credits at current pricing. Its tokens work for this connector only and do not open other Wireflow endpoints.

Limit	Value
Renders	2 per minute, 30 per account per day
Video render	Up to 180 seconds, 60 fps and 1920x1080, in either orientation
Image composite	Up to 4096x4096, and up to 10 rows in a batch
Frame previews	Up to 5 frames per call, 3 calls per minute, 200 still renders per account per day

Add it the same way as the main server, with this URL and a name such as wireflow-editor. In Claude Code, for example:

claude mcp add --transport http wireflow-editor https://www.wireflow.ai/api/mcp/editor

See Editor Connector for every tool.

Add the public skills (optional)

The public skills repo, github.com/wireflowINC/skills, adds nine ready-made skills on top of the connector, such as AI image generation, AI video generation, product photography and a runner for your own workflows. Each skill plans the job, shows the credit cost and your balance, waits for your yes, runs a published Wireflow app and saves the results to ./wireflow-outputs/.

Install them with the skills CLI, which works with Claude Code, Cursor, Codex, Gemini CLI and more:

npx skills add wireflowINC/skills

In Claude Code you can also use the plugin from Option 1 above. See Claude Code Skills.

Test it

Start a new chat with your agent and paste this prompt. It only reads your account, so it spends no credits:

Use Wireflow. Tell me which Wireflow account you are connected as, my credit balance and plan, and list my five most recent workflows.

Your agent should call whoami, get_credit_balance and list_workflows. If it cannot find the Wireflow tools, check that the server is turned on and that you finished the sign-in.

Then try a first generation. This one spends credits, so the prompt asks for the cost first:

Use Wireflow to make a 16:9 image of a red bicycle against a blue tiled wall with Nano Banana 2. Tell me the credit cost and wait for my yes before you run it.

What it costs

API and MCP access are on every plan, including Free. Runs still use credits, with the same credits and model access as your account.

Plan	Price	Credits each month
Free	$0	0
Starter	$24/mo	1,600
Professional	$45/mo	4,300
Team	$249/mo	21,500

On Free, your first image is on us when you finish setup, and video models are on paid plans. Nano Banana 2, used in the test above, costs 20 credits per image. See the pricing page.

MCP or API key?

Use the MCP connector for agents: it signs in with OAuth and needs no key. Use an API key, which starts with sk-, for your own scripts and servers; see API Overview. The two are separate. An MCP sign-in is not an API key: REST endpoints such as the video composition contract accept an sk- key only. API keys and the connector also resolve identities independently, so if you run more than one, have your agent call whoami, or call GET /api/v1/me with the key, before it writes anything.