English | 中文 | العربية | Español
- Overview
- Highlights
- Tour
- What an Engagement Produces
- Quick Start
- A Typical Day
- Persona Packs
- License
- Support & Feedback
FundaPod is a research workbench for portfolio managers. You hire analysts — generic, or persona packs like Warren Buffett, Charlie Munger, and Ray Dalio — assign them tickers, and ask them to write the work a buy-side analyst would: pitch memos, earnings reactions, maintenance updates, and free-form theme explorations. Every claim cites a primary source you can click into.
The workbench is the surface — a browser UI where chats, coverage trees, memos, and a live knowledge graph sit side by side. Everything an analyst produces is a plain file on disk: open it in your editor, grep it, version it, share it. There's no hidden database.
- 🎓 Onboard anyone as an analyst — Hire from bundled persona packs (Buffett, Munger, Dalio) or onboard a new mind by pointing us at a public figure's writings, interviews, or a book. Each onboarded persona becomes a hireable analyst with their own voice and lens.
- ⚗️ Distillation pipeline — Feed in a Wikipedia page, a stack of shareholder letters, or a book; out comes a structured analyst skill — voice, mental models, default workflow — ready for you to refine before putting them on the desk.
- 👥 Run a pod, not a single agent — You're the PM. Hire equity analysts, risk managers, data scientists, data engineers, sector specialists, and more. Each seat keeps its own coverage list, default workflow, and writing voice.
- 🧠 Knowledge graph as your second brain — Memos, tickers, themes, analysts, and citations rendered as one connected board. See what your pod has written, where claims trace back to, and where the gaps are.
- 💡 Idea research across every source — Synthesize across past memos from every seat on the desk, academic papers (arXiv, SSRN, Semantic Scholar), sell-side reports, online content, and the web. The master agent surfaces ideas the pod already holds and hunts for new ones.
- 🛠️ Workflow control per deliverable — Compose the chain of skills behind each output: pitch memos, morning briefs, earnings reactions, maintenance updates, theme explorations. Drop in new skills, reorder steps, or build entirely new memo types.
🎓 Talent Pool — Distilled personas and onboarded analysts, ready to hire.
🧠 Second Brain — Knowledge graph of every memo, ticker, theme, analyst, and citation your pod has produced.
🧰 Skill Library — Atomic skills your analysts chain into workflows — drop in new ones without code.
🧭 Workflow Library — Templated workflows behind each deliverable: pitch memo, earnings reaction, morning brief — fully composable.
🗄️ Data Library — Pluggable data sources available to every seat on the desk.
When an analyst works a ticker, everything lands under data/engagements/<analyst>/<TICKER>/:
| Artifact | Location | Description | |
|---|---|---|---|
| 📄 | Memos | memos/ |
Pitch memos, earnings reactions, maintenance updates, idea write-ups |
| 📚 | Filings | corpus/filings/<FORM>/<ACCESSION>/ |
10-K, 10-Q, 8-K — fetched as clean Markdown via edgartools |
| 📈 | Market snapshots | corpus/snapshots/yahoo/ |
Daily price, 52-week range, analyst consensus, financials |
| 📰 | News & press | corpus/news/, corpus/press/ |
Recent news and press releases |
| 🎤 | Transcripts | corpus/transcripts/ |
Earnings call transcripts when available |
| 🔬 | Research | corpus/research/ |
Web search and academic literature survey notes |
| 📐 | Analysis | analysis/kpis/, analysis/sections/ |
Extracted KPIs and drafted memo sections |
| 🧾 | Coverage brief | .pipeline/docs/coverage_brief.json |
The analyst's living one-pager about this name |
Theme engagements (free-form trading ideas through the master chat) land under a synthetic house/IDEA-<slug>/ so they don't pollute real coverage trees.
- Python 3.10+
- A Claude Code subscription — FundaPod authenticates through Claude Code's OAuth, so there's no separate API key to manage.
- Node.js is not required — the web UI is pre-built and shipped with the package.
git clone https://github.com/<your-username>/fundapod
cd fundapod
pip install -e .npm install -g @anthropic-ai/claude-code
claude /loginFollow the OAuth prompt. FundaPod picks up the credentials automatically.
SEC requires a name + email in the User-Agent for filings requests. Copy .env.example to .env and set:
FUNDAPOD_SEC_USER_NAME=Your Name
FUNDAPOD_SEC_USER_EMAIL=you@example.comfundapod serveOpen http://127.0.0.1:8001 in your browser. From here you can hire analysts, build a watchlist, and run engagements without touching the terminal again.
Prefer the CLI?
A few useful commands if you'd rather drive things from the terminal:
fundapod templates # list available memo workflows
fundapod plan NVDA pitch-memo # plan an engagement (stamp out tasks.json)
fundapod run NVDA pitch-memo # plan + execute end-to-end
fundapod status NVDA # see brief + per-task status
fundapod engagements # list materialized engagements
fundapod universe --sector Technology # browse the US ticker catalogueWorking on the frontend?
The shipped web UI is pre-built, so end users don't need Node. If you want to edit the React app, run a Vite dev server alongside fundapod serve:
cd web
npm install # first time only
npm run dev # Vite on http://localhost:5173, proxies /api to :8001Keep fundapod serve running in another terminal — the dev server forwards /api/* calls to it. When you're done editing, npm run build re-generates the bundle into fundapod/static/ (which is what production fundapod serve ships).
- Pick names. Open My Universe, search the US ticker catalogue, and add the names you care about to your watchlist.
- Hire your team. Drop in a persona pack (Buffett, Munger, Dalio) or distill a new one from a Wikipedia page. Each analyst gets a desk, a voice, and a default workflow.
- Open a chat. Ask Maria Chen (or Warren) "write a pitch memo on NVDA." The right rail shows you the work happening live — filings being pulled, news being read, sections being drafted.
- Read the memo. Every claim is a clickable citation back to the underlying filing, transcript, or news article. Disagree with a take? Reply in the chat and the analyst will rework it.
- Theme work in the master chat. When you want to think across the book — "where are we exposed if the Fed is on hold through Q3?" — the master chat runs a survey and pulls together a two-section memo: which of your existing ideas are exposed, plus new ideas to consider.
FundaPod ships with three bundled investor personas you can hire immediately:
| Persona | Style | Built-in lens | |
|---|---|---|---|
| 🟦 | Warren Buffett | Owner-mindset, moat-first, long-duration | Economic moats, owner earnings, management quality |
| 🟧 | Charlie Munger | Latticework of mental models, inversion | Multidisciplinary checklist, "what would make this a terrible idea?" |
| 🟪 | Ray Dalio | Macro, principles-driven, regime-aware | Big cycles, debt dynamics, regime shifts |
You can also distill a new persona from a public figure's Wikipedia page — FundaPod uses the bundled Buffett skill as a shape template and asks Claude to write the rest. Treat the output as a starting point, then refine by hand.
PolyForm Noncommercial 1.0.0. Free to use, modify, and share for personal projects, research, education, and other noncommercial purposes. Commercial use is not permitted without a separate license — reach out if you'd like to discuss one.
FundaPod is actively under development. The core workflow — hire, watchlist, pitch memo, earnings reaction, theme exploration — works end-to-end. Expect rough edges, especially around non-US names and esoteric data sources.
- 🐛 Found a bug? Open an issue on GitHub.
- 💡 Have an idea or a workflow you wish existed? Open a discussion — feedback shapes the roadmap.
- 📬 Commercial licensing or partnership inquiries? Reach out via the repository contact.