@getaide/sdk

SDK, CLI, and MCP Server for AIDE - Extract structured data from documents using AI.

Installation

# As SDK dependency
npm install @getaide/sdk

# As global CLI
npm install -g @getaide/sdk

# Run without installing
npx @getaide/sdk --help

SDK Usage

import { AideClient, JwtAuth, ApiKeyAuth } from '@getaide/sdk';

// Set AIDE_BASE_URL env var or pass baseUrl explicitly
// e.g. AIDE_BASE_URL=https://aide.datatera.ai

// API key auth (limited to published endpoints)
const client = new AideClient({
  baseUrl: 'https://aide.datatera.ai',
  auth: new ApiKeyAuth('your-api-key'),
});

// JWT auth (full access)
const jwt = await JwtAuth.login('https://aide.datatera.ai', '[email protected]', 'password');
const client = new AideClient({
  baseUrl: 'https://aide.datatera.ai',
  auth: jwt,
});

// Upload and process a file
const result = await client.published.processAndWait(
  'transformation-id',
  fileBlob,
  'invoice.pdf',
);

// List projects
const projects = await client.projects.list(userId);

// Get transformation data
const data = await client.transformations.getData('transformation-id');

CLI Usage

# Login
aide auth login --email [email protected] --password secret
aide auth login --api-key your-api-key

# Projects
aide projects list
aide projects create "My Project"

# Transformations
aide transformations list
aide tdata <transformation-id>

# Process files
aide process upload invoice.pdf -t <transformation-id>
aide pstatus <job-id>
aide presults <job-id>

# History
aide history list -t <transformation-id>
aide history download <history-id> -f excel

# JSON output
aide projects list --json

MCP Server

Use with Claude Desktop, Cursor, VS Code, or any MCP-compatible AI assistant.

Claude Desktop / Cursor

Add to your MCP settings:

{
  "mcpServers": {
    "datatera": {
      "command": "npx",
      "args": ["-y", "@getaide/sdk", "--mcp"],
      "env": {
        "AIDE_API_KEY": "your-api-key",
        "AIDE_BASE_URL": "https://aide.datatera.ai"
      }
    }
  }
}

Or with the dedicated binary:

{
  "mcpServers": {
    "datatera": {
      "command": "aide-mcp",
      "env": {
        "AIDE_JWT_TOKEN": "your-jwt-token",
        "AIDE_USER_ID": "your-user-id",
        "AIDE_BASE_URL": "https://aide.datatera.ai"
      }
    }
  }
}

Environment Variables

| Variable | Description | |----------|-------------| | AIDE_API_KEY | API key authentication | | AIDE_JWT_TOKEN | JWT token authentication | | AIDE_USER_ID | User ID (required with JWT for some tools) | | AIDE_BASE_URL | API base URL (default: https://aide.datatera.ai) | | AIDE_PRODUCT_NAME | Custom product name for white-label (default: AIDE) |

Available MCP Tools

| Tool | Description | |------|-------------| | list_projects | List all projects | | get_project | Get project details with transformations | | create_project | Create a new project | | list_transformations | List transformations | | create_transformation | Create a transformation | | get_transformation_data | Get extracted data | | upload_and_process | Upload and process a file | | get_job_status | Check processing status | | list_history | List processing history | | download_result | Get result details | | list_datamarts | List datamarts | | run_datamart | Execute datamart query | | get_datamart_data | Get datamart results | | run_agent | Trigger DQ agent or Decision Board | | get_agent_status | Check agent status | | list_agent_templates | List agent templates | | smart_project_upload | Upload files for Smart Project | | smart_project_status | Check Smart Project status |

White-Label Support

The SDK supports custom branding via environment variables:

AIDE_BASE_URL=https://app.aidataextractor.pro
AIDE_PRODUCT_NAME="AI Data Extractor"

Works with any AIDE-powered instance including:

• aide.datatera.ai (default)
• app.aidataextractor.pro (Russian market)
• Custom white-label deployments

License

MIT