Give your agent a wallet you control.
Let your agents pay online using one-time-use cards or with machine payment protocols. You approve every request.
Know within minutes whether a secondaries offer deserves an NDA, with a price view from your own fund data, and see which pensions may sell.
FinanceGitHub @alexandrecela10OnlineOwner unconfirmed
Have you tried it? One tap is enough. No sign-in needed.
Repository information has not been collected yet.
No updates detected yet. GitHub releases will appear here.
Secondaries Radar: triages secondaries offers in minutes with a price view from your own fund data, and finds pensions that may sell. Every number sourced. Prototype.
Taken from the repository README
Loading the full README.
Read the full README ↗A working prototype for a fund that buys private equity stakes from investors who want out early.
quicklook/, formerly Quick Look): reads a broker's sale offer, identifies each fund from its numbers (names are hidden before an NDA), and gives a sourced verdict: look closer or pass. Persona and journey: outputs/journey-maps/quick-look-associate-journey.md.firstcall/, formerly First Call): reads real public news and lists pensions showing signs they may sell, with quotes and links. Persona and journey: outputs/journey-maps/first-call-partner-journey.md.Our own portfolio and offers are synthetic, with invented names. Seller Watch uses public news headlines and short quotes only. Not investment advice, and not a statement of any pension's intent.
cd quicklook && make setup && make journey
cd firstcall && make setup && make demo
Optional keys go in .env at the repo root: ANTHROPIC_API_KEY (model parsing and classification), and LANGFUSE_PUBLIC_KEY, LANGFUSE_SECRET_KEY and LANGFUSE_HOST (tracing and experiments). Without keys, both features run on offline parsers and tracing is switched off.
A guided Streamlit demo, written to you: three tabs, one per persona plus "Under the hood". It needs no key and costs nothing.
uv venv --python 3.11 .venv && uv pip install --python .venv/bin/python -r requirements.txt
.venv/bin/streamlit run app.py
.venv/bin/pytest tests # smoke test: data behind each tab, then the app headless
demo/data/firstcall_snapshot.json), with real pension names, headlines, short quotes and article links. All of it is public news; we never store the article itself.python demo/build_snapshot.py [--firstcall-db firstcall/data/firstcall.duckdb] rebuilds demo/data/ (the synthetic universe takes about 10 seconds to generate, so the app reads a copy).Deploy on Streamlit Community Cloud: point a new app at this repo, branch main, main file app.py, and pick Python 3.11 under Advanced settings. requirements.txt installs Streamlit and the two local packages (./quicklook, ./firstcall). Optional live mode: add ANTHROPIC_API_KEY under the app's Secrets. Claude can then read pasted offers and headlines, capped at 3 calls per visitor and 30 per day. Without the secret, live mode is hidden. Never commit .streamlit/secrets.toml.
docs/ARCHITECTURE.mdoutputs/opportunity-analyses/, outputs/prds/, outputs/decisions/