io.github.echomindr/echomindr
Real founder decisions, lessons and signals from 100+ podcasts — searchable by AI agents.
Open source Repository Open in the app JSON README (API)
About
Real founder decisions, lessons and signals from 100+ podcasts — searchable by AI agents.
Details
- Kind
- MCP servers
- Topic
- Social & content
- Publisher
- echomindr
- Origin
- official
- Category
- ferramentas
- Transport
- sse
- Version
- 1.0.0
- Stars
- 1
- Last push
- 2026-03-13T14:58:25Z
- Repository state
- arquivado
- Language
- Python
- License
- MIT
- Added
- 2026-08-29 03:02:44
- Updated
- 2026-08-29 03:02:44
- Origin id
io.github.echomindr/echomindr
README
# Echomindr
**3,500+ real founder moments from 60+ podcasts — searchable by AI agents.**
Each moment: a named founder, a verbatim quote, a decision taken, an outcome observed, a lesson extracted — with a timestamped link to the source. Not summaries. Not paraphrases. What actually happened.
## Why
AI agents give generic startup advice. Echomindr gives them access to what founders actually did.
Ask: *"How did founders handle their first pricing?"*
Get: Kevin Hale's 10-5-20 rule, Josh Pigford charging $249/month from day one, Madhavan Ramanujam's options trick — with quotes, outcomes, and source links.
Ask: *"What did founders do when they nearly ran out of money?"*
Get: Airbnb selling cereal boxes, Notion's near-collapse during COVID, Calm's years of slow growth before the breakout — directly from the founders who lived it.
## Quick start
### API (REST)
```bash
# Search for founder experiences
curl "https://echomindr.com/search?q=pricing&limit=5"
# Describe a situation, get matching experiences (vector search)
curl -X POST "https://echomindr.com/situation" \
-H "Content-Type: application/json" \
-d '{"situation": "B2B SaaS founder with free pilots that won'\''t convert to paid"}'
# Get moment details
curl "https://echomindr.com/moments/{id}"
# Find similar moments
curl "https://echomindr.com/similar/{id}?limit=5"
```
API docs: [echomindr.com/docs](https://echomindr.com/docs)
### MCP (for AI agents)
Connect via remote MCP: `https://echomindr.com/mcp/`
Or add to Claude Desktop (`claude_desktop_config.json`):
```json
{
"mcpServers": {
"echomindr": {
"command": "python",
"args": ["echomindr_mcp.py"],
"env": {
"ECHOMINDR_API_URL": "https://echomindr.com"
}
}
}
}
```
3 MCP tools:
- `search_experience` — semantic search for founder stories by situation
- `get_experience_detail` — full details of a moment (quote, decision, outcome, lesson)
- `find_similar_experiences` — related founder stories by shared themes
### llms.txt
```
https://echomindr.com/llms.txt
```
## Data
- **3,500+ moments** from 340+ podcast episodes across 60+ shows
- **52 canonical situations** across 10 thematic families (PMF, growth, pricing, fundraising, team, operations, resilience, strategy, founder psychology, hostile environments)
- **5 moment types:** decision, problem, lesson, signal, advice
- **5 stages:** idea, mvp, traction, scale, mature
- Sources: How I Built This, Lenny's Podcast, 20 Minute VC, Acquired, Y Combinator, My First Million, GDIY (Génération Do It Yourself), Disrupting Japan, Silicon Carne, Startup Ministerio, Kevin Kamis, Wall Street Paper, Valy Sy (China), Matt & Ari (Canada), Oscar Lindhardt (Denmark), Aidan Walsh (USA)
Each moment: summary · verbatim quote · decision · outcome · lesson · stage · tags · timestamp link
## Self-hosting
To run your own instance with the sample data:
```bash
git clone https://github.com/echomindr/echomindr.git
cd echomindr
pip install -r requirements.txt
# Build a sample database
python echomindr_build_db.py --sample
# Start the API
python echomindr_api.py
# → http://localhost:8000/docs
```
To build the full database, you need your own podcast transcriptions and Claude API key. See `echomindr_extract_v2.py` for the extraction pipeline.
## Architecture
```
Podcast audio → Deepgram (transcription) → Claude (extraction) → SQLite → FastAPI → MCP
```
The extraction pipeline turns long-form podcast interviews into structured, searchable moments. Each episode yields 8–15 moments on average. Semantic search uses BAAI/BGE-M3 embeddings (1024-dim) via sqlite-vec.
## Endpoints
| Endpoint | Method | Description |
|---|---|---|
| `/search` | GET | Full-text search with stage/type filters |
| `/situation` | POST | Describe a situation, get matching experiences (vector search) |
| `/moments/{id}` | GET | Full moment detail |
| `/similar/{id}` | GET | Similar moments by shared tags |
| `/taxonomy` | GET | 52 canonical situations across 10 families |
| `/stats` | GET | Database statistics |
| `/llms.txt` | GET | LLM-optimized API description |
| `/docs` | GET | Swagger documentation |
## License
MIT — the code is open source. The hosted database at echomindr.com is a managed service.
---
Built by [Thierry](https://www.linkedin.com/in/thierryfaucher/) — author of "The System That Learns Wins" and "Designing for Permanent Hostility".