Skip to main content
Log InStart Free

Developers

Build on a platform designed to be extended.

Call it, script it, and add your own connectors, nodes and tools. Each surface below says whether it is available, in preview or planned.

Developer DocumentationBrowse the platform
Ask a deployed model for a prediction· Available in MLDoo (Preview)
# Each deployment has its own endpoint and bearer token.
curl -X POST "$DOOHORIZON_URL/api/ml/serve/$DEPLOYMENT_ID/predict" \
  -H "Authorization: Bearer $DEPLOYMENT_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{ "rows": [
        { "tenure_months": 14, "plan": "pro", "tickets_90d": 3 }
      ] }'

# The same deployment can explain its answer:
#   POST /api/ml/serve/$DEPLOYMENT_ID/explain
# and describe itself:
#   GET  /api/ml/serve/$DEPLOYMENT_ID/schema

The developer platform

Eleven ways in.

No surface is described as shipped unless it is. Planned items are the direction, not documentation.

  • API

    Preview

    A versioned HTTP API with a consistent error envelope and a request id on every response.

    • Projects, data sources, datasets, dashboards and exports under /api/v1
    • Model serving: predict, explain, health, metadata and schema per deployment
    • A public health endpoint and a token-protected metrics endpoint

  • Authentication

    Available

    People and machines authenticate differently, on purpose.

    • Signed, httpOnly session cookies for interactive use
    • A bearer token per model deployment
    • A device token for each paired desktop agent, revocable from the browser

  • SDKs

    Preview

    MLDoo ships a plugin SDK today; language SDKs for the wider platform are planned.

    • Package a custom MLDoo component with typed ports and parameters
    • Plugins are reviewed and approved before a flow can use them
    • TypeScript and Python client SDKs are planned

  • CLI

    Preview

    The desktop agent has a command line today; a platform CLI is planned.

    • Pair a computer, share a folder and check status from a terminal; the agent window says when an update is available
    • Run the agent headless and install it to start at login
    • A CLI for deploying and promoting platform assets is planned

  • Webhooks

    Available

    Start workflows from events in other systems, and tell other systems what happened.

    • A Webhook Trigger gives each flow its own address; callers sign the body with HMAC-SHA256
    • Per-flow rate limits, with Retry-After when a sender loops
    • Notify posts to Slack, Microsoft Teams or any chat webhook; REST API Output calls any endpoint

  • Connectors

    Available

    First-party connectors today, with a public connector interface on the roadmap.

    • Nine databases (MySQL, SQL Server, PostgreSQL, Azure SQL, Oracle, MongoDB, Snowflake, BigQuery, DuckDB), S3, SharePoint, FTP/SFTP, spreadsheets, cloud drives, email, CRM and student-information systems
    • Any REST API as a pipeline step
    • A published interface for third-party connectors is planned

  • Custom nodes

    Preview

    Extend the canvas with your own steps.

    • MLDoo plugins add components to the model designer
    • A SQL step and formulas cover custom logic in FlowDoo today
    • Custom FlowDoo nodes are planned

  • Visualization extensions

    Planned

    Add chart types to VizzDoo beyond the built-in renderers.

    • A declarative chart specification
    • Server-side rendering inside the existing sandbox

  • AI tools

    Planned

    Expose your own actions to AIDoo agents with typed inputs and outputs.

    • Workflows and deployed models as tools
    • Per-tool permissions and approval rules

  • Marketplace

    Planned

    Publish connectors, templates, models, agents and components with clear attribution.

    • First- and third-party assets side by side
    • Review before listing
    • Commercial terms will be published before the marketplace opens

  • Local development

    Available

    Run the whole platform on your own machine.

    • Node.js 20.9 or later; one command starts a development server
    • Everything is configured with environment variables
    • Type-check, lint, unit and end-to-end test suites are included

Code

What it looks like.

The first sample runs against MLDoo today. The second is conceptual: it shows the developer experience we are designing toward, and is labelled so.

Ask a deployed model for a prediction· Available in MLDoo (Preview)
# Each deployment has its own endpoint and bearer token.
curl -X POST "$DOOHORIZON_URL/api/ml/serve/$DEPLOYMENT_ID/predict" \
  -H "Authorization: Bearer $DEPLOYMENT_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{ "rows": [
        { "tenure_months": 14, "plan": "pro", "tickets_90d": 3 }
      ] }'

# The same deployment can explain its answer:
#   POST /api/ml/serve/$DEPLOYMENT_ID/explain
# and describe itself:
#   GET  /api/ml/serve/$DEPLOYMENT_ID/schema

Start building.

Open the documentation, start free, and tell us what you want the platform to expose next.

Start FreeBook Enterprise Demo