Claude Code
ChatGPT
VS CodeE-commerce
Product shots from one photo
“Take this product photo and generate four e-commerce shots: on white, on a marble counter, held in hand, and a lifestyle shot in a bright kitchen. Keep the label and colors exactly as in the original.”
Segmind MCP server is live
Generate media
from any AI agent.
Add the Segmind MCP server to Claude Code, ChatGPT, Codex, VS Code or any MCP client.
500+ image, video and audio models. Sign in with your Segmind account, no API key to paste.
Connector URLpaste into any MCP client
https://mcp.segmind.com/mcpStreamable HTTPOAuth sign-in500+ modelsPixelflowsNo API key
Works with
- Claude Code
- ChatGPT
- Codex
- VS Code
Connect in minutes
Pick your agent.
Paste one URL, sign in, create.
Segmind in Claude Code
One command in your terminal, then authenticate from inside a session.
- 1
Add the server
Registers Segmind as an HTTP MCP server. Add --scope user to make it available in every project.
Terminalclaude mcp add --transport http segmind https://mcp.segmind.com/mcp - 2
Authenticate
Start Claude Code, run the /mcp command, pick Segmind and choose Authenticate. Sign in to Segmind in the browser that opens.
Inside Claude Code/mcp - 3
Start creating
Ask for an image, a video or a voiceover. Claude Code browses the catalog, checks the price and runs the model.
How it works
You describe it.
Your agent finds the model and runs it.
The server exposes the live catalog as tools. The agent browses it, checks what a run costs, submits the job and hands you the output.
agent session · segmind mcp
@segmindMake a hero image for our coffee brand: a ceramic cup on a sunlit wooden table, steam rising, editorial look.
list_models({ search: "text to image" })compare_models({ slugs: ["nano-banana-pro", "gpt-image-2"] })run_inference({ slug: "nano-banana-pro", params: { prompt: "…", aspect_ratio: "16:9" } })
Output readynano-banana-pro16:92K28 s$0.15
Here is the hero image. Want a portrait crop for social as well?
Prompts to start with
Say it the way you would
to a designer.
Copy a prompt into your agent. It picks the model, fills in the parameters and returns the result.
Marketing
Campaign variations
“Turn this ad still into three campaign variations for spring, summer and holiday. Change the setting and props, keep the product, framing and the tagline placement the same.”
Video
Animate a still
“Animate this sneaker photo into a 5-second 9:16 clip: slow orbit around the shoe on a wet street at night, neon reflections, subtle camera push-in at the end.”
Audio
Voiceover for a script
“Read this 40-second product explainer as a warm, confident narrator. Natural pacing, a short pause after each sentence, and return an MP3 I can drop into the edit.”
Editing
Edit with instructions
“Remove the background from this headshot, replace it with a soft studio grey, fix the harsh shadow under the chin and export at 2048 px on the long edge.”
Concept art
Storyboard a scene
“Storyboard a six-panel sequence of a courier cycling through Tokyo at dawn: wide establishing shot, close-up on the handlebars, a near miss with a taxi, and a final rooftop reveal. Consistent character, cinematic 2.39:1 frames.”
Every model, inside your agent
The models developers run most
on Segmind this month.
The server reads the live catalog, so every model on Segmind is available the day it ships. These are the ones with the most active accounts in the last 30 days.
The tools
13 tools.
Find it, price it, run it, collect it.
The agent discovers models at runtime rather than carrying a tool per model. One catalog, one set of verbs, whatever you are making.
Find a model
The catalog is read live, so a model is callable the day it ships and nothing here goes stale.
list_modelsBrowse and rank the catalog by cost, latency or popularity.compare_modelsPut a shortlist side by side with their real price tables.get_model_specThe full spec: parameters, pricing, examples.
Generate
Images come back inline on the fast path. Video and other slow models submit and collect, so nothing times out.
run_inferenceRun a model and wait for it. The fast path for images.submit_inferenceStart a slow job and return its id straight away.check_inferenceCollect the output once the job finishes.
Run a Pixelflow
Pixelflows are saved multi-model workflows. Your agent runs a whole pipeline instead of chaining models itself.
list_pixelflowsBrowse published workflows and the models they chain.get_pixelflow_specThe inputs a flow takes and the models it runs.submit_pixelflowStart a flow run.check_pixelflowCollect the flow's outputs.
Inputs and account
Attach a photo in the chat and the agent can use it as a model input. It can also check the balance before an expensive job.
upload_mediaTurn an attached file into a URL a model can read.get_creditsThe signed-in account's credit balance.pingHealth check and server version.
Pricing
Your Segmind credits. Nothing extra.
The connector calls the same API you would call yourself, on your own account. Each generation costs exactly what the model page lists.
Same price as the API
No connector fee
Billed per generation
Balance via get_credits
FAQ
Questions, answered
A hosted Model Context Protocol server that gives AI agents access to Segmind's model catalog. Your agent can browse models, compare what they cost, run a generation, run a saved workflow and check your credit balance, all from the chat you are already in.
Claude Code, ChatGPT in Developer mode, Codex, and VS Code 1.106 or later with Copilot. Any other client works too, as long as it supports remote MCP servers over Streamable HTTP and signs in using a Client ID Metadata Document.
Some clients only support dynamic client registration, which Segmind's authorization server does not offer, so their sign-in cannot complete yet. That is the only thing standing in the way, and support for more clients is being added.
No. Add the server URL and sign in with your Segmind account when the client asks. The connector acts on your account, so there is no key to copy, store or rotate.
Anything in the Segmind catalog: images from text or reference images, image editing, image-to-video and text-to-video, text to speech, music and sound effects, upscaling and background removal. Over 500 models are available, and new ones appear automatically because the server reads the live catalog.
Each generation uses your existing Segmind credits at the same price as calling the API directly. There is no extra charge for using the connector, and your agent can compare model prices before it runs anything and check your balance before a long job.
Images usually return in a few seconds and are shown inline. Videos take one to several minutes, so the agent submits the job, hands you the request id and collects the result when it is ready rather than blocking the chat.
Yes, in clients that pass attachments to tools. Attach the image to the chat and the agent uploads it to a hosted URL the model can read. Note that the URL is public to anyone holding it, which is what lets the model provider fetch it.
Saved multi-model workflows published on Segmind, such as a product photo pipeline or a try-on flow. Your agent can browse them and run one end to end instead of chaining models itself. A flow bills every model in its chain.
Yes. The Connected Apps page in your Segmind console lists every app you have authorized and lets you revoke it at any time. The client stops working immediately until you sign in again.
Available now
One URL.
Every media model, in your agent.
Paste the connector URL, sign in with your Segmind account, and start creating from the chat you already use.
New to Segmind? Create a free account