What Does It Do?

Purpose

AI Foundation is shared infrastructure. It does not replace your CMS — it makes AI reliable inside TYPO3.

Child extensions and MCP agents send requests to AI Foundation. AI Foundation routes them to your configured AI provider or to T3Planet Credits. The provider (OpenAI, Claude, Gemini, Mistral, and others) returns the result. Every call is logged.

Your extensions / MCP agents
         ↓
   AI Foundation (ns_t3af)
         ↓
   AI provider OR T3Planet Credits
         ↓
   OpenAI · Claude · Gemini · Mistral · …

1. AI Providers

Path: AI Foundation → AI Providers

Connect to major AI vendors from one backend screen.

  • Add multiple providers (OpenAI, Anthropic, Gemini, Azure, Mistral, DeepSeek, xAI, custom, Ollama)

  • Store API keys encrypted

  • Pick models and run Test connection

  • Set one default provider for the whole system

See AI Providers for step-by-step setup.

2. MCP Server

Path: AI Foundation → MCP Server

Expose TYPO3 to external AI assistants via the Model Context Protocol.

  • Remote OAuth — production (Cursor, Claude Desktop)

  • mcp-remote — URL token for simple clients

  • Local CLI — DDEV and development

Core tools: read pages, list content, read/write database records (with permissions).

See MCP Server.

3. AI Context

Path: AI Foundation → AI Context

Store brand identity once: company name, audience, tone, keywords. Injected into AI prompts automatically for on-brand output.

See AI Context.

4. AI Prompts

Path: AI Foundation → AI Prompts

Central library of prompt templates. Extensions can sync and reuse them.

See AI Prompts.

5. AI Features

Path: AI Foundation → AI Features

Assign different providers per task type (SEO, pages, content, translation).

See AI Features.

6. Usage & Logs

Path: AI Foundation → AI Usage / AI Logs / Scheduler & CLI

  • AI Usage — charts, tokens, credits

  • AI Logs — per-request detail for audits

  • Scheduler & CLI — background jobs

See AI Usage & Logs.

7. AI Permissions

Path: AI Foundation → AI Permissions

Control who uses which provider, which capabilities, budgets, rate limits, and log privacy. Off by default — enable when your team grows.

See AI Permissions.

Supported providers

  • OpenAI — Chat, vision, embeddings, images

  • Claude / Anthropic — Long-form text, analysis

  • Google Gemini — Fast multimodal tasks

  • Azure OpenAI — Enterprise deployments

  • Mistral — EU-friendly option

  • DeepSeek — Cost-effective chat

  • xAI — Grok models

  • Custom / Ollama — Private or local models

Public API (for developers)

Child extensions call AiServiceInterface only:

  • complete() — Full AI text response

  • stream() — Real-time text chunks

  • embed() — Vector embeddings

  • provider() — Active provider record

Extensions never call vendor APIs directly — AI Foundation handles keys, logging, credits, and governance.

Full reference: Developer Guide

Typical admin workflow

  1. Install extension → Installation

  2. Add provider and run Test connection → AI Providers

  3. Run Quick Setup → Quick Setup

  4. (Optional) Enable MCP → MCP Server

  5. (Optional) Set governance rules → AI Permissions

  6. Monitor usage weekly → AI Usage & Logs

Requirements at a glance

  • TYPO3 — 12.4 LTS, 13.4 LTS, or 14.x

  • PHP — 8.2 or higher (8.3 recommended)

  • System extensions — workspaces and scheduler

  • Billing — API key or T3Planet Credits (at least one)

  • Composer — Recommended for TYPO3 installs

See System Requirements for full details.