Build on Aleria, under the same rules.

# one API for every model, inside the country
curl https://gateway.your-org.aleria.ai/v1/chat/completions \
  -H "Authorization: Bearer $ALERIA_KEY" \
  -d '{
    "model": "aleria-reason",
    "messages": [{"role": "user",
      "content": "Summarise the Q3 risks"}]
  }'

from aleria import Aleria
client = Aleria()   # reads ALERIA_KEY
pack = client.experts.run(
    expert="finance-analyst",
    project="board-q3",
    task="Prepare the Q3 board pack",
)
print(pack.status)   # waiting_for_approval

# install once
curl -fsSL https://aleria.ai/install.sh | sh
# run an expert in a project
aleria run "board pack for Q3" --project board
# ✓ finance analyst · 4 systems read
# ✓ Q3 board pack.pdf · waiting for your approval

// any MCP client: point it at your Aleria
{
  "mcpServers": {
    "aleria": {
      "url": "https://mcp.your-org.aleria.ai",
      "auth": "oauth"
    }
  }
}

# start an automation from your own system: its Webhook trigger gives the URL
curl -X POST "$WEBHOOK_URL" \
  -H "x-api-key: $ALERIA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{ "event": "contract.signed", "id": "C-2041" }'

Endpoint names and SDK calls are illustrative until the docs are public.