Connect Eden Gen to your AI agent

Pick your agent, copy one prompt, paste it. Then ask for images and video in the chat you already use. Every take is priced first and filed to your project.

Read the full setup docs

Paste this into Claude

Set up Eden Gen so I can create images and videos here. It is a remote MCP server at https://edengen.ai/mcp that signs in with OAuth. If you are Claude Code, run `claude mcp add --transport http eden-gen https://edengen.ai/mcp`, then tell me to open /mcp and finish the browser sign-in. If you are the Claude app, walk me through adding it as a custom connector, one step at a time. Once it is connected, call takes_account and takes_list_models. Tell me which workspace is connected, my balance and monthly cap, what I can make, and what to try first.

In Claude Code it adds the server itself. In the Claude app it walks you through Connectors, Add custom connector. Then it opens a browser tab where you sign in once and pick a project and a monthly cap.

Other ways to connect

1In Claude Code: run the command below, then type /mcp, pick eden-gen and sign in in the browser
2In the Claude app: open Connectors (under Customize or Settings), click Add, choose Add custom connector and paste the MCP URL
3Approve a project and a monthly cap, then ask for your first take

Open Claude connectors

Custom connectors in the Claude app need a paid Claude plan. On Team and Enterprise an owner adds them first.

Read the full setup docs Any client that speaks Streamable HTTP MCP works too.

One chat, from brief to file.

Real asks, real tool calls. Every image and video in this window came out of Eden Gen MCP, priced before it ran.

All of Eden Gen inside the chat.

MCP brings the whole workspace to your agent, not just a render button.

Every model

157 image, video and audio models. Name one, or let your agent pick for the shot.

Best-value route

Each take is priced on every route that is ready and runs on the best value.

Price before it runs

Your agent quotes the exact price first. Above your threshold it stops and asks.

Caps per connection

Every connection belongs to one project with a monthly cap. Nothing runs past it.

Files in your projects

Every take lands in its project with the prompt, model, route and cost.

Refunds on failure

If a provider fails, the take is refunded on the spot, with the reason.

Say what you want. Eden Gen makes it.

Ask in your own words. Your agent picks the model, writes the prompt and brings back the result.

Two 4:5 key visuals for a coffee brand: espresso in amber backlight, and a candlelit table.

Espresso pouring into a glass cup, crema swirl, warm backlight making the coffee glow amber, black background, droplets frozen

Overhead rustic table with roasted vegetables, bread, red wine, warm candlelight, dark moody styling

Images

Art-direct stills in plain words

Key visuals, product shots, a whole campaign. Your agent can call Nano Banana Pro, GPT Image 2.5 or Nano Banana 2.1, and start from your own references.

Make an image

A surfer inside a teal barrel, slow motion, then the same light on a kayak between icebergs.

A surfer inside a glassy teal barrel, slow motion

The kayaker paddles between cobalt icebergs, from a Nano Banana Pro still

Video

Direct the shot, sound included

Describe the camera, the light and the pace, or start from a still. Seedance 2.5, Kling 3.0 and Veo 3.1 render it, priced before it runs.

Make a video

Use our dancer element under one orange spotlight, then the same light on a sprinter for the sequel.

Contemporary dancer mid-leap in a dark theater, single orange spotlight, flowing fabric, dust in the beam, frozen motion

Silhouette of a sprinter mid-stride against a tangerine sky, low angle, sweat backlit, high contrast, Nike-style ad photography

Elements

Keep a character in every shot

Save a face, a product or a location once as an element. The agent passes it as a reference, so the hero looks the same from the first frame to the last.

Create an element

Render all twelve shots of the storyboard in 1080p.

12 takes on Seedance 2.5 at 1080p, 10 s each

$67.50

Above your $5 threshold. Run it?

Yes, renderDraft in 480p first

Budgets

It asks before it spends

Every connection has a project and a monthly cap. Above your threshold the agent stops, shows the price and waits for your yes.

Set a cap

MCP or API? Same engine, two ways in.

MCP

Eden Gen comes to your chat

For people working in Claude, ChatGPT, Cursor, Codex or VS Code.

You are writing a brief and want three hero variants right there
Your IDE is where the work happens
Your team's assistant should render without anyone leaving it
You want a price check before every take

Connect your agent

API

Eden Gen runs in your pipeline

For scripts, apps and batch jobs, with REST and OpenAI-compatible endpoints.

The same flow runs every week
A brief should return 200 variants, not one
Takes start from a webhook or a cron job
Results go straight to your own bucket

Read the API docs

Questions, answered before you connect.

What is MCP?

The Model Context Protocol lets an AI assistant use outside tools. Connect Eden Gen once and your agent can list models, quote a price, render and hand back the file.

Which agents work with Eden Gen MCP?

Claude, Claude Code, ChatGPT, Cursor, Codex, VS Code, Gemini CLI and any client that speaks Streamable HTTP MCP. Clients with MCP sign-in connect through the browser; the rest use an API key.

Do I need an API key?

Not for clients with MCP sign-in. You approve the project and a monthly cap on a consent screen. Other clients use a key from API keys, scoped to one project.

How does pricing work?

The price of the best-value route plus your plan fee, the same as in the workspace, shown before the take runs. Failed takes are refunded to your balance. Above your threshold the agent asks before it renders.

How do I know which model and route ran?

Every reply and every take page shows the model, the route that ran it and the price. The same record is in your usage export.

Where do my results go?

Into the project the connection belongs to, with the prompt, model, cost and route, next to everything your team made in the workspace.

What is the difference between MCP and the API?

Same engine, same prices. MCP is for working in a chat or an IDE. The REST and OpenAI-compatible API is for your own scripts, pipelines and apps.

Is my data used to train AI models?

We do not train models on your prompts or outputs, and we never sell them. Requests go to the provider APIs under their API terms.

Bring Eden Gen into your agent.

Connect now