MCP Server
Connect Claude, Cursor or any Model Context Protocol client to PixelPanda. Ask the agent to remove a background, upscale a render or generate product photos, and it calls the tools directly — hosted endpoint with OAuth, or a local install.
Let the Agent Do the Image Work
An AI agent that can write your listing copy should also be able to produce the photo that goes with it. MCP is the standard that makes that possible, and PixelPanda speaks it.
Point any MCP client at the hosted endpoint and the agent gains a toolbox: background removal, upscaling, enhancement, text removal, prompt-based editing, image-to-prompt, product photo and scene generation, avatars, try-on and UGC video. Authentication is OAuth in the browser or a pk_live_ API key, and every call is metered against the same credit balance as the web app.
What the agent can do
- Process — remove backgrounds, upscale 2x-8x, enhance, unblur, remove text, convert formats.
- Generate — product photos, lifestyle scenes, avatars, virtual try-on and UGC video.
- Reverse — image_to_prompt turns any image into prompts for Flux, Midjourney or Stable Diffusion.
- Free local tools — crop, resize and format conversion run on the machine with no credits at all.
What Agents Do With It
The image half of an agentic workflow.
Hosted endpoint
Point a client at pixelpanda.ai/mcp — no local install and no API key pasting.
OAuth or API key
Full OAuth with PKCE, or a pk_live_ bearer token if you prefer.
33 tools
Processing, generation, reverse prompting and free local editing in one server.
Same credits
Agent calls spend the same balance as the web app, with per-call accounting.
Local option
pip install pixelpanda-mcp to run it on your own machine over stdio.
In the official registry
Listed at registry.modelcontextprotocol.io so clients can discover it.
Hosted, OAuth, no install
One command connects a client — or pip install pixelpanda-mcp to run it locally.
How It Works
One command, then just ask.
Add the server
claude mcp add --transport http pixelpanda https://pixelpanda.ai/mcp — or pip install pixelpanda-mcp for local stdio.
Authorise
Complete the OAuth prompt in the browser, or set PIXELPANDA_API_TOKEN for the local server.
Ask for an image task
Tell the agent what you want — remove a background, upscale, generate product photos.
Get results back
The agent receives hosted result URLs and reports credits used and remaining.
Who Uses the MCP Server
People whose assistant should finish the job, not describe it.
Claude & Cursor users
Ask the assistant to clean up or generate an image as part of a longer task, with no context switch.
Agent builders
Give a custom agent real image capability through a standard protocol instead of bespoke API glue.
E-commerce operators
Have the agent take a supplier photo, cut it out and generate listing images in one conversation.
Developers
Prototype image pipelines conversationally before committing them to code.
Content teams
Turn a still and a script into a UGC-style video by asking for it.
Anyone with a folder of photos
Point the agent at the files and let it batch process them with the free local tools.
Frequently Asked Questions
How do I connect Claude to PixelPanda?
What is MCP?
Can I run it locally instead?
Which tools does the agent get?
How is it authenticated?
How much does it cost?
Are there other PixelPanda MCP servers?
Connect PixelPanda to your agent
Hosted MCP at pixelpanda.ai/mcp, listed in the official MCP registry.
Explore More Resources
Browse our guides, tutorials, and tools