TurfAI DevelopersSources & credits
The external references whose foundational explanations and diagrams we adapted, with credit.
This documentation is TurfAI's own. For the foundational explanations of widely-shared AI concepts — tool use, the agent loop, retrieval-augmented generation, multi-agent orchestration — we've adapted framing, a few worked examples, and some diagrams from the authoritative sources below, and credited them here and inline on the page. Everything specific to the TurfAI platform (its APIs, task types, configuration, and behavior) is original.
Where a concept goes deeper than the foundations, we link out to the original source rather than reproduce it — those vendors maintain the canonical, most current treatment.
What we adapted
| Borrowed | Source | Used as | Where |
|---|---|---|---|
| The "augmented LLM" building block | Anthropic — Building effective agents | Diagram + framing | Agents |
| Orchestrator–workers pattern | Anthropic — Building effective agents | Diagram + framing | Squads |
| Prompt-chaining & routing workflows | Anthropic — Building effective agents | Diagrams | Workflows |
| Autonomous-agent loop; agent-vs-workflow guidance | Anthropic — Building effective agents | Diagram + framing | AI-app patterns |
| The tool-calling loop | OpenAI — Function calling | Diagram + framing | Integrations |
| Grounding / retrieval with citations | Google — Gemini API grounding | Diagram + framing | Knowledge base & RAG |
| The "smallest citable unit" idea | Anthropic — Search results | Framing | Knowledge base & RAG |
| MCP as "a USB-C port for AI" | Model Context Protocol | Diagram + framing | Integrations |
| Durable prompt-engineering tactics | Anthropic, OpenAI, Google | Framing | Prompts |
| Gating actions by reversibility & blast radius | Anthropic — Claude prompting best practices | Framing | Human-in-the-loop |
Diagrams are © their respective owners (Anthropic, OpenAI, Google) and are reproduced here for
educational reference with attribution. Trademarks belong to their owners. The machine-readable
record of every borrowed item lives in docs-sources.json at the repository root and is checked
for upstream changes via the docs-sync routine.