MCP server

Connect Pedra to Claude, ChatGPT, Cursor, and any MCP client. The official Model Context Protocol server exposes one tool per Pedra endpoint — virtual staging, renovation, image enhancement, and listing videos — as blocking calls that return the final asset URL.

Official MCP server

Open source, MIT-licensed, and published to npm as @pedra-ai/mcp. It exposes one tool per API endpoint; each tool is a single blocking call that returns the final asset URL(s) — no job IDs to poll. Under the hood it's a thin wrapper over the Pedra Node SDK.

Use it in Claude (web, mobile & desktop)

In Claude, open Settings → Connectors → Add custom connector, paste the URL above, and click Connect — you sign in to Pedra once and approve access. No API key to copy and paste. It works in Claude on web, mobile, and desktop, and runs on your account's credits.

What you can do

Once connected, ask your assistant in plain language — it picks the right tool and returns the finished image or video URL:

"Use Pedra to virtually stage https://example.com/empty-living-room.jpg as a minimalist living room."

"Empty this furnished bedroom so I can show the bare room: https://example.com/bedroom.jpg"

"Renovate this dated kitchen in a Scandinavian style: https://example.com/kitchen.jpg"

"Add a large green plant in the corner of this living room: https://example.com/living.jpg"

"Brighten and straighten the photo I just attached."

"Swap the grey sky in this exterior shot for clear blue: https://example.com/facade.jpg"

"Blur the faces and license plates in this street-facing photo: https://example.com/street.jpg"

"Make a listing video from these 8 photos with music and a '3 bed · 2 bath' caption."

"How many Pedra credits do I have left?"

Working with your photos

Photos already in your Pedra account are the easiest source — just ask, e.g. "build a video from my Calle Mayor listing photos", and the assistant lists your projects, reads their image URLs, and renders the video. To pull in a photo from the web, give its URL and it's added to a project (the server fetches it for you). For brand-new photos on your device, the assistant hands you a link to open the project in Pedra and drop them in — chat can't upload large files directly.

Other clients & local setup (Cursor, scripts, Claude Desktop config)

For Claude, prefer the hosted connector above. For Cursor, local scripts, or editing Claude Desktop's config file directly, run the npm package over stdio — it launches with npx (no global install) and reads your key from the PEDRA_API_KEY environment variable.

{
  "mcpServers": {
    "pedra": {
      "command": "npx",
      "args": ["-y", "@pedra-ai/mcp"],
      "env": { "PEDRA_API_KEY": "your-api-key" }
    }
  }
}

In Claude Desktop, open Settings → Developer → Edit Config and add the pedra entry. In Cursor, use ~/.cursor/mcp.json (or .cursor/mcp.json in a project). Restart the client after saving.

Authentication

The server reads your key from PEDRA_API_KEY. Get one by signing up at app.pedra.ai and opening Settings → API. See Authentication for details.

Tools

One tool per endpoint, all returning final URLs:

• pedra_furnish, pedra_empty_room, pedra_renovation, pedra_edit_via_prompt, pedra_remove_object — see Virtual staging.
• pedra_enhance, pedra_enhance_and_correct_perspective, pedra_sky_blue, pedra_blur — see Image enhancement.
• pedra_create_video, pedra_update_video, pedra_generate_voice_script, pedra_generate_voice, pedra_music_library, plus pedra_credits and pedra_feedback — see Video and Credits.
• pedra_list_projects, pedra_list_project_images, pedra_create_project, pedra_add_images_to_project — browse your library and add photos so any tool can use them.

Most image tools take an imageUrl — a public URL, a data: URI, or an absolute path to a local file (read and inlined for you). Plus a few optional parameters — your client shows each tool's input schema. Note: an image pasted into the chat isn't sent to the tool — drag in a file or pass its path. Every image and video tool also accepts an optional propertyId: pass it and the result is saved into that property's gallery — visible and editable from the app — instead of only coming back as a URL. See Properties.

How it works

A thin wrapper over the Pedra Node SDK. Every endpoint blocks and returns the final URL(s) inline — even pedra_create_video, which can take up to ~10 minutes (the API keeps the connection alive with a heartbeat). API errors — insufficient credits, a bad image — come back as readable MCP tool errors, not crashes.

Source & support

The server is MIT-licensed and developed in the open at github.com/pedra-ai/pedra-mcp. Learn more about the protocol at modelcontextprotocol.io, and email felix@pedra.ai for integration help, higher quotas, or enterprise terms.

Frequently asked questions