Give your agent a render tool, not a design tool.
Connect ImageClaw as an MCP server and any compatible agent can generate on-brand visuals mid-task, without leaving the conversation or opening a canvas editor.
{"mcpServers": {"imageclaw": {"url": "https://mcp.imageclaw.dev/sse","auth": { "apiKey": "$IMAGECLAW_API_KEY" }}}}
For one banner, card, or OG image at a time.
Renders a hero image sized for a blog post, using the workspace's default blog template family.
| Param | Type |
|---|---|
| title | string |
| category | string |
| accent_color | string, optional |
{ "title": "...", "category": "marketing" }→ { url, width: 1200, height: 630 }
Renders a platform-sized card , LinkedIn, X, Instagram , picking dimensions automatically from the platform name.
| Param | Type |
|---|---|
| platform | string |
| title | string |
| theme | string |
| brand | brandKitId |
{ "platform": "linkedin", "title": "..." }→ { url, width: 1080, height: 1080 }
Renders an Open Graph preview image for a page , used automatically when publishing to most CMSs.
| Param | Type |
|---|---|
| title | string |
| description | string |
{ "title": "...", "description": "..." }→ { url, width: 1200, height: 630 }
For a whole document, generated as one visual set.
A blog post with five images shouldn't look like five unrelated brand-colored PNGs. These tools lock a style once, then keep every image within it as the agent writes.
Locks icon set, illustration style, and layout family for this document. Returns a documentId.
Given the draft or outline, returns which sections warrant an image and why.
Renders within the locked style, returns embeddable markdown and alt text.
Works with any MCP client.
Tested and supported with the major agent runtimes people actually build with.