Minalyst is available for acquisition — a working multi-agent system with backend, frontend, brand, and domain, built and documented end-to-end. I'm moving on, so it's ready for an owner who can ship it.
A multi-agent system that produces structured investment research on public companies. Users ask in natural language — "write a primer on NVDA" — and the system orchestrates a Supervisor, Planner, DataRetriever, WebSearcher, ScreeningAgent, MainAnalyst, and a SEC-filing navigator to assemble a sectioned document with citations, tables, and charts, streamed live into an in-browser editor with PDF and Word export.
FastAPI + LangGraph multi-agent system. ~21,400 LOC across 60 files, 169 pytest tests. SEC EDGAR ingestion, hybrid RAG over Vertex AI Vector Search + Firestore, a financial data layer (FMP / Polygon / Alpha Vantage with Redis caching), a screener with a filter DSL, OAuth, JWT auth, and a streaming agent runtime.
Next.js 16 / React 19 / TypeScript. Streaming agent chat with rich content rendering, a TipTap document editor with PDF and Word export, a screener UI, an AEO-optimized MDX blog (17 posts), share-and-clone with anonymous users, and LemonSqueezy checkout. Builds clean: 0 TS errors, 0 lint errors.
The minalyst.com domain transfers on close, alongside the brand kit, all published blog content, and the public LLM crawl files. Keep the brand and SEO surface, or rebrand with the documented one-pass recipe.
At a conservative $200k/year loaded engineering cost, rebuilding this from scratch runs roughly $306,000 — the base agent system ($204k), the EDGAR pipeline ($68k), and the hybrid RAG layer ($34k) — plus 6+ months of calendar time before you have a head start.
This is an asset sale, priced as a fraction of rebuild cost. Think of it as the cost-avoidance number you put in front of your CFO, not a valuation.
Parser, chunker, validator, and an on-the-fly filing navigator with grep-style and read-by-line-range access over SEC filings under a token budget. No library gets you 80% of the way there.
Supervisor + Planner + ClarificationAnalyzer + DataRetriever + WebSearcher + ScreeningAgent + MainAnalyst + FilingNavigator. The value is in the prompts, routing, loop detection, and recursion tuning — hardened against real traffic.
Embedding service, vector store, Firestore chunk store, and document processor. Vertex AI Vector Search is quick to demo and slow to operationalize — index lifecycle, namespaces, dedup, and GCS staging are all handled.
SSE-driven streaming with reconnect, a document editor that renders agent-generated sections, tables, and charts, plus PDF (Puppeteer) and Word (docx) export — composed with the agent pause/resume protocol.
Honest disclaimers: no revenue worth modeling (one $9/mo user, refundable or transferable), code-only with no production data or API accounts transferred, and a documented set of known issues. This is engineering value, brand, and SEO surface — sold transparently.
No NDA required — this is engineering, not patent-worthy IP. Send one line about who you are and what you have in mind, and I'll share the full teardown deck and doc set.