founders-archive-mcp
# Founders Transcript Archive
A local-first system for building, verifying, searching, and querying a private
archive of [Founders podcast](https://www.founderspodcast.com/) transcripts.
It combines a resumable archive pipeline with full-text and semantic retrieval,
a command-line interface, an MCP server, and a Codex research skill.
## Publication boundary
This repository contains the software, tests, and operating documentation. It
does **not** contain podcast transcripts, indexes, audio, API credentials, or
other generated archive data. Those artifacts live under the Git-ignored
`data/` directory and must remain private. Users are responsible for obtaining
and using source material in accordance with the applicable terms and rights.
## Capabilities
- Discover the canonical episode catalog from the public Megaphone RSS feed.
- Import transcripts through an explicitly capped, resumable Spoken workflow.
- Fall back to local Apple Silicon transcription when a transcript is absent.
- Store portable Markdown and JSON artifacts with a SQLite status/search index.
- Verify catalog coverage, checksums, and artifact integrity.
- Build a cited passage index with hybrid FTS5 and semantic retrieval.
- Query through the CLI or the `founders-archive-mcp` stdio server.
## Requirements
- Python 3.12 or newer
- [uv](https://docs.astral.sh/uv/)
- Optional: Apple Silicon and the `local` dependency group for local
transcription
## Setup
```sh
git clone https://github.com/jacksonxbxt/founders-transcript-archive.git
cd founders-transcript-archive
uv sync --extra research
uv run --extra research --with pytest pytest -q
```
Preview an archive sync without consuming transcript credits:
```sh
uv run --extra local founders-archive sync --dry-run --archive-root data
```
Build the private research index and run a cited query:
```sh
uv run --extra research founders-archive index --archive-root data
uv run --extra research founders-archive research \
"maintaining control while financing growth" \
--archive-root data --limit 8 --json
```
See [PROJECT_PLAYBOOK.md](PROJECT_PLAYBOOK.md) for the complete operating
workflow and safety constraints.
## License
No open-source license has been selected. Public visibility does not grant a
license to copy, modify, or redistribute the code or any underlying content.
TDQS
Scored across 4 tools
Each tool has a clearly distinct purpose: status check, search for evidence, expand a passage, and locate an episode. No overlap or ambiguity between them.
All tools follow a consistent verb_noun snake_case pattern (search_founders, get_founders_passage, find_founders_episode) with the exception of founders_archive_status which uses a noun, but it still reads clearly and the pattern is uniform.
Four tools is well-scoped for this server's purpose. Each one earns its place, covering the essential research workflows without bloat.
The server covers status, search, passage expansion, and episode lookup, which are sufficient for core research tasks. A minor gap is the lack of a browse/list function for episodes, but the existing tools handle the primary use cases.