MCP Integration

Connect any AI agent to Tenkai vision tools via Model Context Protocol.

⚙ "USB-C for AI Agents"

MCP is an open protocol that standardizes how AI applications connect to external tools. Tenkai exposes all vision capabilities as MCP tools — any compatible agent can use them natively.

✦ tenkai_image_restoration

Restore and upscale degraded images using HYPIR. Supports denoising, artifact removal, and 1-8x super-resolution in a single pass.

✨ tenkai_generate_image

Generate images from text prompts using FLUX.2 [klein] 4B. Custom dimensions, steps, and guidance scale.

✏ tenkai_edit_image

Edit an existing image using a text prompt with FLUX.2 [klein] 4B. Accepts image URL or base64 data URI.

↑ tenkai_video_super_resolution

Upscale video resolution 2x or 4x using SeedVR2 one-step diffusion. Supports up to 121 frames per clip.

▶ tenkai_video_generation

Generate videos from text or images using LTX-Video 2B. Text-to-video and image-to-video, up to 10 seconds at 24fps.

⬢ tenkai_segmentation

Segment objects in images by open-vocabulary text prompt using SAM3. Returns masks, bounding boxes, and confidence scores.

MCP Tool Definition

JSON — MCP tools/list response
// Tenkai MCP Server exposes 6 tools { "tools": [{ "name": "tenkai_image_restoration", "description": "Restore and upscale images using HYPIR", "inputSchema": { "type": "object", "properties": { "image": { "type": "string" }, "upscale": { "type": "integer", "default": 4 } }, "required": ["image"] } }, ...] }

Connect — Claude Desktop

Add to claude_desktop_config.json:

Claude Desktop — Settings > Developer > Edit Config
{ "mcpServers": { "tenkai": { "url": "https://tenkai.click/mcp" } } }

Connect — Cursor / VS Code

Add to .cursor/mcp.json or .vscode/mcp.json:

Cursor / VS Code — MCP Config
{ "mcpServers": { "tenkai": { "url": "https://tenkai.click/mcp" } } }

Connect — Python

Python — FastMCP Client
from fastmcp import Client # Connect to the remote Tenkai MCP server async with Client("https://tenkai.click/mcp") as client: # List available tools tools = await client.list_tools() # Upscale an image result = await client.call_tool( "tenkai_super_resolution", {"image_url": "https://example.com/photo.jpg", "scale": 4} ) print(result)

Agent Workflow Example

1
User: "Restore this old photo and upscale it to 4x"
2
Agent calls tenkai_image_restoration → restores degradation + upscales 4x
3
Agent returns the enhanced image with before/after comparison
4
User: "Now generate a video of this scene coming to life"
5
Agent calls tenkai_video_generation with the restored image → image-to-video
6
Agent returns a 10-second video at 24fps

Live Agent Demo

See how an AI agent uses Tenkai MCP tools in a real conversation:

User
Restore this old family photo and make it high resolution
Agent
I'll restore and upscale the photo using HYPIR.
⚙ tenkai_image_restoration(image="photo.jpg", upscale=4)
Agent
Restoration complete! The image has been denoised, artifacts removed, and upscaled to 4x resolution. Here's your restored photo.
User
Can you also segment the people in the photo?
Agent
I'll detect and segment the people.
⚙ tenkai_segmentation(image="restored.jpg", text_prompt="person")
Agent
Found 3 people in the photo with confidence scores 0.95, 0.92, 0.88. Individual masks and an overlay image are ready.

This is the power of MCP — AI agents that autonomously see, evaluate, and enhance visuals.

Supported Clients

Claude Desktop, VS Code Copilot, OpenAI Agents SDK, Cursor, Windsurf, and any MCP-compatible AI host.

Transport

Supports both stdio (local) and Streamable HTTP (remote) transports. Hosted at https://tenkai.click/mcp.

Open Standard

Built on MCP spec. JSON-RPC 2.0 wire format. Image results returned as CDN URLs.

Create your account

Get started with Tenkai API Suite

Already have an account? Sign in