MCP · Model Context Protocol

Kampana MCPfor any AI agent

Every visual interaction in Kampana - browsing apps, building canvases, running fashion workflows, managing products, assets, materials, AI models and Brand DNA - is exposed as an MCP tool. Connect Claude, Cursor, or any MCP client and let agents run your fashion stack.

Streamable-HTTP MCP endpointhttps://app.kampana.io/api/mcp
Works with
Claude DesktopClaude CodeCursorChatGPTCodexPerplexityClineContinueZedWindsurfOpen WebUICustom MCP clients
CAPABILITIES

What an agent can do inside Kampana.

Browse & run fashion apps

Agents can list every workflow in your workspace - Collection Concept, Tech Pack, PDP Pack, B2B Showroom, Marketplace Feeds, Social Campaign - and trigger full runs with the inputs you'd type in the UI.

Create & wire canvas nodes

Spin up new canvases programmatically: add product nodes, fabric swatches, mood references, AI model nodes, and connect them into the same node graph you'd build by hand.

Manage your libraries

Read and write Products, Assets, Materials, AI Models, Brand DNA, and Custom Nodes. The agent uses the same library your team uses - no shadow copies, no drift.

Brand DNA-aware generations

Agents pull your Brand DNA (palette, tone, model preferences, channel rules) before generating a single image or line of copy, so output ships on-brand by default.

Generate visuals & video

Trigger product images, lifestyle scenes, flat lays, front/back lockups, and short-form video directly through the same generation pipeline the canvas uses.

Scoped, auditable access

Authenticate with a workspace API key. The agent can only do what the key's scopes allow, only within that workspace. Every action is logged.

SETUP

Three steps. No SDK.

01

Generate a workspace API key

In Kampana, open Settings → API keys and create a key scoped to the workspace you want the agent to act in. You choose read-only or read/write.

Settings → API keys → Create key
02

Add the MCP server to your client

Point Claude Desktop, Cursor, or any MCP client at the Streamable-HTTP endpoint and paste your API key as the bearer token.

{
  "mcpServers": {
    "kampana": {
      "url": "https://app.kampana.io/api/mcp",
      "headers": { "Authorization": "Bearer kmp_live_..." }
    }
  }
}
03

Talk to your workspace

Say things like "Build a PDP pack for the new linen shirt" or "Refresh the Paris carryover with autumn lifestyle visuals." The agent reads your libraries, builds the canvas, runs it, and returns assets.

> Generate a 6-image PDP pack for product
  "linen-shirt-ss26" using our Brand DNA.
AGENT-NATIVE FASHION OS

If you can click it in Kampana, an agent can call it.

Most fashion-AI tools are click-only. Kampana was built node-first, so every visual interaction already has a typed contract behind it. We just exposed those contracts as MCP tools. The agent doesn't simulate Kampana - it uses the same APIs your UI uses.

  • Same product library, same Brand DNA, same canvases.
  • Run a workflow from a single sentence in Claude or Cursor.
  • Streamable-HTTP transport - no local proxy, no Node bridge.
  • Scoped API keys per workspace, full audit log on every call.
> list_workflows()
→ collection_concept, tech_pack, pdp_pack, b2b_showroom, marketplace_feeds, social_campaign, carryover_refresh, end_to_end_launch …
> create_canvas(workflow="pdp_pack", product="linen-shirt-ss26")
→ canvas_id=cv_8fJ2 · nodes=[product, brand_dna, model_pool, image_gen×6]
> run_canvas("cv_8fJ2")
→ 6 images · 1 hero · 1 flat-lay · status=ready
> list_assets(product="linen-shirt-ss26")
→ 12 assets returned
Workspace-scoped keys
Each key is bound to one workspace. Rotate or revoke any time.
Granular permissions
Read-only, generation, library write - choose what the agent can touch.
No data exfiltration
Agents act inside Kampana. Generated assets stay in your workspace, not the client.
Agent-ready

Plug Kampana into your AI workflow.

Free workspaces include MCP access. Bring your favorite agent - keep the Kampana canvas as your source of truth.