SVG Lab is an MCP server with 90 tools an agent can call, among them account_status, add_artboard, add_hidden_ui. It answers at https://svglab.app/mcp/agents over streamable HTTP, and it listed its tools to an anonymous client, so no sign-in was needed for that. Our probes reached it 4 of 4 times since 25 September 2026; the last check, on 8 November 2026, succeeded.
Design engine for AI: app screens, icons, illustrations and charts as clean SVG in SVG Lab projects.
Description as the publisher wrote it.
| Tool | What it does |
|---|---|
account_status | Report the plan, access, SVG Lab credits left and what actions cost. |
add_artboard | Add a screen, slide, page or post (preset sizes it). What a control opens (dropdown, filter, menu) is add_hidden_ui. Inputs: width, height, x, y, name, background |
add_hidden_ui | Use when a control opens hidden UI: adds a sized, linked dropdown, side, sheet, dialog or state board beside its screen. Inputs: screen, trigger, kind, to, name, height |
add_texture | Add hatch, stipple, facet or strata marks clipped to a shape of a drawing on its "Drawing · <illustration>" board. On a screen the marks are refused as a picture drawn onto it. Inputs: shape_id, kind, angle, spacing, jitter, seed |
align | Align shapes to a shared edge or centre. Inputs: shape_ids, mode |
align_to_artboard | Position shapes relative to their artboard. Inputs: shape_ids, horizontal, vertical |
apply_layout_fixes | Move misaligned layout children to their suggested positions. Inputs: target, artboard, group, shape_ids, min_font_size, ignore_pairs |
bind_slot | Fit artwork into a layout slot and keep it bound to it. Inputs: shape_ids, layout_id, path, mode, clip, detach |
boolean_op | Union, subtract, intersect or exclude shapes into one new path. Inputs: op, shape_ids |
check_contacts | Measure whether parts touch what they belong to. Inputs: pairs, attachments, tolerance, artboard, display_width |
check_contrast | Compute WCAG contrast between two colours. Inputs: foreground, background, target, font_size, font_weight |
clip_shapes | Clip or mask shapes with another shape. Inputs: target_ids, clip_id, as_mask |
clip_to_artboard | Clip shapes that cross their artboard's frame. Inputs: artboard, shape_ids, include_clipped |
compare_drawing | Measure your drawing against the picked image: overlay, difference map, pixels off and the shape misses left; candidates score edits before you send them. measure reads the picked image itself (a gridded crop, colour samples, colour runs). Inputs: artboard, reference_id, region, regions, candidates, ignore |
create_icon | Draw a built-in monoline icon (docs {components} names them). Inputs: icon, x, y, size, color, stroke_width |
create_layout | Build a screen or section from a layout tree; returns boxes, ids and measured warnings. With a shell (Claude Code): write the fields to a JSON file, curl it to the upload link new_project gave and pass only upload_id, so the user approves a Inputs: verbose, artboard, x, y, width, height |
create_path | Draw one path from anchors with optional absolute handles. With a shell (Claude Code): write the fields to a JSON file, curl it to the upload link new_project gave and pass only upload_id, so the user approves a short call. Inputs: name, style, opacity, style_ref, artboard, points |
create_shapes | Draw up to 80 shapes, text, paths and pencil strokes in one call; illustrations are copied from a reference. With a shell (Claude Code): write the fields to a JSON file, curl it to the upload link new_project gave and pass only upload_id, s Inputs: shapes, artboard, upload_id |
current_project | Show the current project and its link. |
define_styles | Define named style presets. With a shell (Claude Code): write the fields to a JSON file, curl it to the upload link new_project gave and pass only upload_id, so the user approves a short call. Inputs: styles, upload_id |
delete_shapes | Delete shapes. Inputs: shape_ids |
describe_document | Describe the artboards and shapes. Inputs: artboard, name_contains, ids, limit, offset |
design_tokens | Change the project's design tokens. With a shell (Claude Code): write the fields to a JSON file, curl it to the upload link new_project gave and pass only upload_id, so the user approves a short call. Inputs: preset, reset, set, user_named_fonts, imagery, upload_id |
detach_symbol | Turn one symbol instance into editable shapes. Inputs: shape_id |
distribute | Space three or more shapes evenly. Inputs: shape_ids, axis |
docs | SVG Lab's reference: start-here, each tool's full details, design guides and the illustration routine. docs {fonts} lists the type catalog; docs {components} lists the components, icons and layout node types. Inputs: tool, tools, guide, guides, lesson, lessons |
draw_svg | Upload the SVG you wrote by hand of the picked image onto its drawing artboard, with differences: every difference you still see, however small ([] at shape_misses 0). Auto-traced SVG (potrace, vtracer, colour quantising, any script that tu Inputs: artboard, svg, upload_id, name, reference_id, differences |
duplicate | Copy shapes in place or at an offset. Inputs: shape_ids, offset |
edit_path | Change a path's anchors and handles in place. With a shell (Claude Code): write the fields to a JSON file, curl it to the upload link new_project gave and pass only upload_id, so the user approves a short call. Inputs: shape_id, coordinate_space, artboard, subpath_index, move_points, set_handles |
export_artboard | Export an artboard as SVG, with a preview card and a link to a PNG picture. Inputs: artboard, outline_text, background, part, export_id |
export_code | Export screens as code (HTML, React, SwiftUI, Compose, Flutter) or layered SVG, with tokens and assets, as a ZIP link. Inputs: format, artboards, file |
export_document | The finished deck or document by link: PDF, editable PowerPoint (pptx) or post images. Inputs: format, artboards, file_name |
find_references | SVG Lab's reference pictures for one planned illustration (illustration, query): Creative Commons first, each with its licence, no stock-agency previews; candidates are for you only. Inputs: query, style, count, illustration |
fit_artboard | Fit the active artboard to the artwork on it, plus optional padding. Inputs: padding |
fit_to_box | Scale and move shapes together into a box; placing an illustration removes its drawing board. Inputs: shape_ids, x, y, width, height, mode |
fit_to_children | Resize a container to wrap its children. Inputs: container_id, child_ids, padding |
flip | Flip shapes horizontally or vertically. Inputs: shape_ids, axis |
generate_illustration | Draw a planned illustration with SVG Lab's vector model: editable SVG from a text brief, the faster route (500 credits). Inputs: illustration, prompt, artboard |
get_assets | Get drawings and icons as SVG or PNG files by asset id or artboard; a ZIP link for many. Inputs: artboard, assets, format, scales |
group | Group shapes. Inputs: shape_ids, artboard, name |
illustration_jobs, inspect_layout, inspect_path, list_artboards, list_design_styles, list_layouts, list_projects, list_shapes, list_symbols, look, make_responsive, make_symbol, measure_optical_center, measure_text, mirror, name_group, new_project, normalize_radii, normalize_spacing, offset_path, open_project, optical_center, pause_editor, pencil, place_image, place_symbol, plan_illustrations, read_screen, reorder, repeat, resize, resolve_overflow, scatter, search_photos, set_active_artboard, set_artboard, set_filter, set_gradient, set_locked, set_pattern, set_reference_image, set_style, set_text, smooth_path, snap_to_grid, text_on_path, transform, ungroup, update_layout, upload_asset
Every block below points at https://svglab.app/mcp/agents. Pick your client and copy.
claude mcp add --transport http svg-lab https://svglab.app/mcp/agents
{
"mcpServers": {
"svg-lab": {
"command": "npx",
"args": ["-y", "mcp-remote", "https://svglab.app/mcp/agents"]
}
}
}{
"mcpServers": {
"svg-lab": {
"url": "https://svglab.app/mcp/agents"
}
}
}{
"servers": {
"svg-lab": {
"type": "http",
"url": "https://svglab.app/mcp/agents"
}
}
}{
"context_servers": {
"svg-lab": {
"url": "https://svglab.app/mcp/agents"
}
}
}{
"mcpServers": {
"svg-lab": {
"serverUrl": "https://svglab.app/mcp/agents"
}
}
}{
"mcpServers": {
"svg-lab": {
"url": "https://svglab.app/mcp/agents"
}
}
}name: SVG Lab
version: 0.0.1
schema: v1
mcpServers:
- name: SVG Lab
type: streamable-http
url: https://svglab.app/mcp/agents{
"mcpServers": {
"svg-lab": {
"httpUrl": "https://svglab.app/mcp/agents"
}
}
}{
"mcpServers": {
"svg-lab": {
"type": "http",
"url": "https://svglab.app/mcp/agents"
}
}
}{
"mcpServers": {
"svg-lab": {
"url": "https://svglab.app/mcp/agents"
}
}
}goose://extension?url=https%3A%2F%2Fsvglab.app%2Fmcp%2Fagents&type=streamable_http&id=svg-lab&name=SVG%20Lab
It answered our last check on 8 November 2026 in 1124 ms. Over 4 checks since 25 September 2026 it answered 4. Status here comes from Neuronto's own probes, not from the publisher.
Yes. It is published by svglab.app and served from svglab.app, the vendor's own domain: https://svglab.app/mcp/agents.
90, as returned by the server's own tools/list on 15 November 2026: account_status, add_artboard, add_hidden_ui, add_texture, align, align_to_artboard, apply_layout_fixes, bind_slot, and 82 more.
It answered our last check on 8 November 2026 in 1124 ms. Over 4 checks since 25 September 2026 it answered 4.
Put the server under mcpServers in ~/.cursor/mcp.json with the URL https://svglab.app/mcp/agents. The exact JSON is above, with the same setup for Claude Code, Claude Desktop, VS Code, Windsurf, Zed, Gemini CLI and the rest.
It listed its tools to an anonymous client, so reading the tool list needs no key. Individual tools can still ask for credentials in their own inputs.
Publisher: app.svglab. Endpoint host: svglab.app. ARD identifier: urn:air:app.svglab:mcp:svglab.
First indexed 25 September 2026. The tools are exactly what the server returned; nothing on this page
is inferred from its name. Ranking in Neuronto search is never for sale.