Focus on Foundations

David Deutsch

Explore David Deutsch’s ideas through decades of interviews, with quoted answers and source transcripts.

AI Q&A (QRAG)

Ask questions across the David Deutsch interview corpus.

Transcripts

Browse interview excerpts and extracted Q&A. Full transcripts are available when your account has access.

Skills, CLI & MCP

Use the Deutsch corpus from an AI agent or the command line with the QRAG toolkit.

Skills

Agent workflows for searching sources and asking questions with QRAG.

CLI

Use the fof-qrag command-line toolkit to search and ask questions about Deutsch.

MCP

Connect QRAG tools to an MCP-compatible AI assistant.

Skills

Download four ready-to-use agent skills for searching, asking questions, reading transcripts, and managing your account. Each folder includes a SKILL.md entry point and setup instructions.

Download agent skills ZIP

Extract the four folders into your agent’s skills directory, such as ~/.claude/skills for Claude Code or ~/.codex/skills for Codex. The skills use the CLI below. If it is already installed, you can also run:

fof-qrag skills install --target ~/.claude/skills
Skills ZIP checksum (SHA256)d115c1ba5b3a38a4e62da8c3ef00e752df72286031f6fbad71420a31ada6e617

CLI

Download the QRAG toolkit 0.1.0 with the command-line tool, Python library, MCP server, and bundled agent skills. Python 3.10 or later is required.

Download toolkit ZIP

With pipx installed, open a terminal in the folder where you saved the ZIP and run:

pipx install "./fof-qrag-0.1.0-501716d9a4b6.zip"

Then review the privacy notice and search the Deutsch corpus:

fof-qrag init
fof-qrag search "Why are good explanations hard to vary?"

Use fof-qrag ask for an AI answer, or fof-qrag --help for all commands. Public search and excerpts work without an account; full transcripts and restricted models require appropriate access.

Toolkit ZIP checksum (SHA256)501716d9a4b64f80cbb0ca922e32fbde9bbdc64e321626aeb1e0152b23a116d3

MCP

The MCP server is included in the toolkit download. After installing the CLI with pipx, add its MCP dependency:

pipx inject fof-qrag "mcp>=1.10,<2"

Run fof-qrag init in a terminal to review the privacy notice before connecting your assistant. For Claude Code:

claude mcp add fof-qrag -- fof-qrag-mcp

For another MCP client, configure fof-qrag-mcp as its server command with stdio transport. The server provides search, AI Q&A, transcript, and conversation tools, using Deutsch by default.