Docs

Technical reference for Deckbase integrations, APIs, and developer guides.

Reference

Quick start

Terminal
curl -X POST https://www.deckbase.co/api/mcp \
  -H "Authorization: Bearer YOUR_ACCESS_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","method":"tools/list","id":1}'

Replace YOUR_ACCESS_TOKEN with an OAuth token from your Connect page. For setup instructions, see the MCP setup guide.

API overview

Deckbase provides two integration paths: the MCP server for AI-assisted deck and card management, and direct REST API endpoints for programmatic access.

  • Create, update, and delete decks and cards in batches
  • Generate AI audio and images for cards via ElevenLabs and image models
  • Export full deck data including block snapshots and metadata
  • Build custom templates with flexible block layouts
  • Receive webhook events for deck and card changes

Start here

Best practices

  • Treat card creation as an operational pipeline

    Keep request logs, validate required template fields before writes, and spot-check generated cards before scaling batch size.

  • Define one canonical template strategy

    For teams, define one canonical template strategy and one retry policy. This avoids duplicate prompt drift and makes incident response much faster when mapper rules change.

  • Follow the recommended reading order

    Setup page first, MCP Server reference second, then production rollout checks. This path minimizes authentication mistakes and schema mismatch errors during card creation.

  • Document rules in your repo

    If you are implementing team workflows, document your template rules and batch acceptance gates directly in your repo so future updates stay consistent.

  • Run a weekly operational review

    Check recent write failures, template drift, and duplicate prompt rates before scaling automation volume.

  • Version your integration checklist

    Keep your integration checklist versioned so onboarding new contributors does not require reverse-engineering previous MCP setup decisions.

  • Include a rollback runbook

    When possible, include a short runbook for rollback and reprocessing failed batches so production updates remain safe even when templates evolve.

FAQ

  • The Model Context Protocol (MCP) server lets AI tools like Cursor, Claude Code, and VS Code read your docs, list decks, and create cards programmatically. It uses standard JSON-RPC over HTTP.

    Read more

For guides and articles, see Resources.