What the MCP connector does
The Model Context Protocol (MCP) lets an AI assistant call outside tools. The Beemm Vision connector gives it 27: list your projects and workflows, check your balance and gallery, generate images (alone or in batches of 12) and video, upscale, remove the background, inpaint (mask by URL) or expand an image, and create, edit and run a workflow. Results land in your Beemm Vision project as if you had made them in the app.
27
tools exposed by the connector (19 read-only)
2 clients
that generate: Claude and ChatGPT; Cursor and Claude Code read
200 / 50
images and clips a day at most, per account
The four steps
Step 1
Copy the connector address
The address is https://app.beemmvision.com/mcp. Nothing to install: it is a remote connector, and you sign in with your Beemm Vision account.
Step 2
Add it to your client
In claude.ai or Claude Desktop: Customize → Connectors, then + Add → Add custom connector. In ChatGPT: chatgpt.com/plugins, + then Add custom MCP server, with OAuth. In Claude Code: the command claude mcp add --transport http beemmvision followed by the address. In Cursor: the address in .cursor/mcp.json, or the “Installer dans Cursor” (Install in Cursor) button in the app. Those two only read for now: projects, workflows, gallery, balance.
Step 3
Approve the access screen
Beemm Vision shows what the assistant will be able to do: read your projects, create and edit your workflows, launch generations that spend your credits. It is all or nothing: Autoriser (Allow) or Refuser (Deny). You can revoke access anytime in Réglages → Applications connectées (Settings → Connected apps; the app is in French).
Step 4
Ask in plain language
“Generate an image of <subject> in my project <name>”, “animate the image you just made”, “run my workflow <name> and show me the result”. Paid tools answer at once with an id the assistant follows to the result. In claude.ai results show as cards; elsewhere, as text and images.
Each client at a glance
| Client | How to add it | What it can do |
|---|---|---|
| claude.ai, Claude Desktop | Customize → Connectors → custom connector | Everything: read, build, generate, edit; cards in claude.ai |
| ChatGPT | chatgpt.com/plugins → Add custom MCP server (OAuth) | Everything: read, build, generate, edit; depends on your ChatGPT plan |
| Claude Code | claude mcp add --transport http beemmvision <address> | Read-only for now: projects, workflows, gallery, balance |
| Cursor | .cursor/mcp.json or the “Installer dans Cursor” button | Read-only for now: projects, workflows, gallery, balance |
| Other MCP clients | Same address | Read-only: projects, workflows, gallery, balance |
Ideas to try
Four variants in one request
“Plan four variants of <subject> on four image models, show me the cost, then run them”: one batch, one card.
Finish an image without leaving the chat
“Upscale the last image 2× without going over 20 credits”: an edit never goes past the cap it is given.
Re-run a workflow
“Duplicate my workflow <name>, run the copy and tell me when it is ready”: 30 nodes at most, 20 of them paid.
See where the credits go
“Show my balance and my last 10 jobs, grouped by model”: free and read-only.
Frequently asked questions
- Can Claude generate images and videos?
- Yes, with an MCP connector. Once Beemm Vision is added to claude.ai or Claude Desktop, Claude calls Beemm Vision to generate images and video with the catalogue models (Kling, Seedance, FLUX, GPT Image…), on your Beemm Vision credits. Claude Code connects read-only for now.
- What about ChatGPT?
- Yes: ChatGPT connects to the same connector and can generate. Adding a custom MCP server depends on your ChatGPT plan and workspace settings. Visual cards only show in claude.ai.
- How much does it cost?
- The connector is free. Generations, edits and workflow runs spend your Beemm Vision credits. Before an image or a batch, a free quote (a card in claude.ai) shows the cost when it can be computed; an edit accepts a credit cap, required for upscaling and fill inpainting; a video or a workflow run has no quote tool. The assistant’s own tokens are billed by your AI provider.
- Can the assistant spend without limits?
- No. You first approve the access screen, and spending is capped per account: 200 images and 50 clips a day, 20 generations per ten-minute window, across all clients. A call retried with the same idempotency key is billed once. The connector works on your personal projects only.
- The connector does not offer to generate: what now?
- It probably stayed read-only. Remove it in Réglages → Applications connectées (Settings → Connected apps), then reconnect it and approve the access screen: granted access never gains rights when it renews. Also check that your client generates: Claude (claude.ai, Desktop) or ChatGPT. Claude Code and Cursor connect read-only for now.
- Do I have to sign in again often?
- No. The connector renews its access on its own: you stay connected as long as you use it at least once every 30 days and do not revoke it. Only opening a workflow in the agent’s browser is limited to two hours.
Connect Beemm Vision to your assistant
Create your account, then paste https://app.beemmvision.com/mcp into Claude or ChatGPT.