Real products built on ALDO AI.

POST a multipart form. Returns text/event-stream NDJSON; the final `data: {"type":"done", ...}` carries the imageUrl.

curl -N -X POST https://ai.aldo.tech/live/picenhancer/api/enhance \
  -F file=@portrait.jpg \
  -F scale=1 \
  -F bg=1 \
  -F weight=0.7
# Then GET https://ai.aldo.tech/live/picenhancer/api/out/<filename> for the PNG.

Streams the SSE; resolves with the final result once the pipeline emits `done`.

async function enhance(file) {
  const fd = new FormData();
  fd.append('file', file);
  fd.append('scale', '1');     // 1 / 4 / 8 / 16
  fd.append('bg', '1');        // 0 = leave background alone
  fd.append('weight', '0.7');  // GFPGAN strength
  const res = await fetch('https://ai.aldo.tech/live/picenhancer/api/enhance', {
    method: 'POST', body: fd,
  });
  const reader = res.body.getReader();
  const decoder = new TextDecoder();
  let buf = '';
  while (true) {
    const { value, done } = await reader.read();
    if (done) break;
    buf += decoder.decode(value, { stream: true });
    let i; while ((i = buf.indexOf('\n\n')) >= 0) {
      const line = buf.slice(0, i).split('\n').find(l => l.startsWith('data:'));
      buf = buf.slice(i + 2);
      if (!line) continue;
      const ev = JSON.parse(line.slice(5).trim());
      if (ev.type === 'done') return ev;          // { imageUrl, faces, scale, ... }
      if (ev.type === 'progress') /* update UI */;
    }
  }
}

Same wire shape; works identically from any HTTP client.

import json, requests
from sseclient import SSEClient

with open('portrait.jpg', 'rb') as f:
    res = requests.post(
        'https://ai.aldo.tech/live/picenhancer/api/enhance',
        files={'file': ('portrait.jpg', f, 'image/jpeg')},
        data={'scale': '1', 'bg': '1', 'weight': '0.7'},
        stream=True,
    )
done = None
for ev in SSEClient(res).events():
    payload = json.loads(ev.data)
    if payload['type'] == 'done':
        done = payload
print(done['imageUrl'], done['faces'], done['enhanceMs'])

Drop into your MCP client config. The server runs as a stdio child process and exposes one tool: `picenhancer.enhance`.

// claude_desktop_config.json (or any MCP client that speaks stdio)
{
  "mcpServers": {
    "picenhancer": {
      "command": "npx",
      "args": ["-y", "@aldo-ai/mcp-picenhancer"]
      // Optional override:
      // "env": { "PICENHANCER_BASE_URL": "https://ai.aldo.tech/live/picenhancer/api" }
    }
  }
}

Once the MCP server is registered, any chat / agent can call:

// tools/call request body
{
  "name": "picenhancer.enhance",
  "arguments": {
    "image": "data:image/jpeg;base64,/9j/4AAQ...",   // or an https:// URL
    "mode": "enhance",                                // enhance | enhance-bg | upscale-x4 | upscale-x8
    "strength": 0.7                                   // GFPGAN weight 0.0–1.0
  }
}
// Response (structuredContent):
//   {
//     imageUrl, scale, bg, weight, faces,
//     origDims, enhancedDims, origBytes, enhancedBytes, enhanceMs
//   }

Wire the MCP tool into any composite agent in your tenant — same pattern as the bundled aldo-fs MCP server.

# agency/your-agent.yaml
name: my-image-agent
tools:
  permissions:
    mcp:
      - picenhancer.enhance:
          allow: ["*"]
prompt: |
  When the user gives you a portrait photo, call picenhancer.enhance
  with mode="enhance" and strength=0.7. Return the resulting imageUrl
  and the face count to the user.