Values and color
Perceptual value bands, dominant colors, exact point sampling, and nearest artist-paint matches.
Model Context Protocol · Image processing
PaintValues lets an AI agent inspect a painting reference, find its value structure and paint colors, then return practical studies, grids, edge maps, or transformed images. Connect through one remote MCP endpoint in the same web app.
01 / Direct answer
The PaintValues MCP server is a remote image-processing service for AI agents working with painting references. It exposes structured tools through the open Model Context Protocol, so a compatible agent can use an attached image or HTTPS image URL without a separate PaintValues application server.
The agent chooses a bounded tool, PaintValues performs the pixel work, and the result returns as structured data, an image, or both. The AI can then explain what the result means for the painting rather than guessing from a text-only description.
A typical request
“Can you help me find the values to paint for this image?”You, with a reference image attached
The agent finds dominant colors, value distribution, and close paint matches.
PaintValues groups perceptual lightness into clear, paintable shapes.
The agent can discuss the largest masses, focal contrast, and a starting palette.
02 / Capabilities
The service covers analysis, study generation, drawing transfer, and practical image preparation. Each operation has a constrained schema so the agent can select settings deliberately.
Perceptual value bands, dominant colors, exact point sampling, and nearest artist-paint matches.
Value studies, color massing, printable reference sheets, and simplified edge drawings.
Count grids, crosses, diagonals, and perspective rays for moving a reference to canvas.
Crop, resize, rotate, flip, blur, recolor, isolate color, and convert to PNG, JPEG, or WebP.
03 / Tool reference
Tools accept a native image attachment or an HTTPS image URL unless the row says otherwise. Generated images are returned inline at a bounded size suitable for MCP clients.
| Tool | What it does | Returns |
|---|---|---|
paintvalues_capabilities | Discover supported formats, limits, presets, retention, and privacy behavior. | Structured service facts |
inspect_painting_reference | Analyze dominant colors, perceptual values, and nearest artist-paint colors. | Structured analysis |
sample_reference_color | Sample an exact point and find its nearest colors in PaintValues palettes. | Color and paint matches |
create_value_study | Group the reference into 3–50 perceptual value bands. | PNG, JPEG, or WebP |
create_painter_reference | Apply a painter preset, adjustments, and optional composition guides. | PNG, JPEG, or WebP |
create_reference_sheet | Combine the original, five-value study, color masses, and edges on one sheet. | Printable image |
create_edge_reference | Find simplified contours for block-in and stroke planning. | Edge image |
create_grid_reference | Add count grids, crosses, diagonals, or perspective rays for transfer. | Gridded image |
isolate_reference_color | Keep a picked Lab color range or dominant RGB channel in color. | Isolation image |
transform_painting_reference | Crop, rotate, flip, resize, blur, recolor, and convert the reference. | PNG, JPEG, or WebP |
delete_uploaded_reference | Permanently delete an owned private PaintValues upload after confirmation. | Deletion result |
04 / Connect
Use the production endpoint as a remote, Streamable HTTP MCP server. Compatible clients can discover the OAuth authorization server from PaintValues metadata and request only the scopes a tool needs.
https://www.paintvalues.com/mcpThe exact settings label varies by client. No local package or second PaintValues server is required.
05 / Inputs and limits
Supported clients can pass their native file object directly to a tool. This is the simplest path for a user-attached reference.
An agent can submit a public HTTPS URL. Private, loopback, link-local, and other unsafe network targets are rejected.
When a client cannot supply either input, use the private upload page. The default retention window is 24 hours, and an owned upload can be deleted sooner.
06 / Privacy and security
Attached and linked images are fetched for the selected tool and held in memory while processing. Optional private uploads expire after the configured retention window. Usage records contain operational facts such as tool name, status, duration, byte counts, and dimensions—not image contents, filenames, URLs, prompts, email addresses, or access tokens.
The public endpoint is an OAuth 2.1 protected resource with separate scopes for processing, private upload, and deletion. The interactive browser editor remains a separate, browser-local workflow.
07 / Common questions
No. The MCP route runs from the same PaintValues Next.js application and Vercel deployment as the web app. Connect to https://www.paintvalues.com/mcp.
Yes, when the MCP client supports native file parameters. Every PaintValues processing tool also accepts an HTTPS image URL.
Yes. An agent can request perceptual value groups, dominant colors, a sampled color, and nearest matches in the PaintValues artist-paint palettes.
Attached and linked inputs are processed in memory. Only the optional private-upload route retains a private blob temporarily; it expires automatically and can be deleted sooner.
Any compatible remote MCP client can use the endpoint. File attachment behavior and the location of MCP settings depend on the client.
PaintValues accepts JPEG, PNG, and WebP images up to 10 MB, with a maximum decoded size of 40 megapixels.
The public endpoint is an OAuth 2.1 protected resource. Compatible clients discover the authorization server from PaintValues metadata and request separate scopes for image processing, private upload, or deletion.
A clearer reference, on request
Connect the MCP server, attach a reference, and ask in ordinary language.