AI ECM Agent Squad for M&A, IPO & Public Markets
liquidround.ai · Predictive Labs Ltd
LiquidRound is an AI-powered chat-first workspace for investment banking and equity capital markets. A squad of specialist AI agents covers the full deal lifecycle — buyer-led sourcing and diligence, seller-led positioning and IPO readiness, and public markets intelligence.
flowchart TB
USER(["User"]) -->|HTTPS| SERVER
subgraph SERVER["FastHTML Process · port 5007"]
LANDING["Landing & Blog\n/, /blog, /agents"]
APP["Chat App\n/app — 3-pane shell\nSSE streaming"]
SCHED["Scheduler\ndigest · candidates\nsecurity returns"]
end
subgraph DATA["PostgreSQL"]
LR[("liquidround\nusers · workflows · messages\nipo_pipeline · deal_candidates")]
HF[("hedgefolio\n13F holdings · 10K funds\n3.4M positions · $71T AUM")]
end
subgraph EXT["External APIs"]
XAI[["Grok LLM\napi.x.ai"]]
TAVILY[["Tavily"]]
EXA[["EXA"]]
YF[["yfinance"]]
PM[["Postmark"]]
end
APP --> LR
APP --> HF
APP --> XAI
APP --> TAVILY
APP --> EXA
APP --> YF
SCHED --> PM
SCHED --> LR
- Buy-side M&A — find acquisition targets, run diligence, score matches, build IC memos
- Sell-side M&A — prepare teasers / CIMs, identify buyers, assess IPO readiness
- Public Markets — IPO tracking (map, pipeline, prospectus), SPAC lifecycle dashboard
- Hedge Fund Intelligence — SEC 13F institutional holdings, fund AUM rankings, activist filings
- BYOD — upload pitch books, CIMs, term sheets; agents read, cite, and draft from them
23 specialist agents across 6 categories. The auto-router picks the right agent from natural language, or use a prefix shortcut (scan:, dcf:, memo:, hedgefunds:, etc.).
flowchart TD
MSG["User Message"] --> ROUTER["Router\nagents/router.py"]
ROUTER -->|"scan: triage:\nintent: buyer:"| SRC["Deal Sourcing"]
ROUTER -->|"comps: ltm: dcf:\nmulti: synergy:"| UND["Valuation"]
ROUTER -->|"vdr: abstract:\nlegal: ops: esg:"| DIG["Due Diligence"]
ROUTER -->|"memo: teaser:\nbid: ipo: score:"| CAP["Capital"]
ROUTER -->|"research:\nintegrate:"| PORT["Portfolio"]
ROUTER -->|"hedgefunds:"| PUB["Public Markets"]
style ROUTER fill:#F59E0B,color:#0B1220,stroke:#D97706
style SRC fill:#3B82F6,color:#fff
style UND fill:#8B5CF6,color:#fff
style DIG fill:#EC4899,color:#fff
style CAP fill:#10B981,color:#fff
style PORT fill:#F97316,color:#fff
style PUB fill:#06B6D4,color:#fff
|
🔍 Deal Sourcing (4)
|
📊 Valuation & Underwriting (6)
|
📋 Due Diligence (5)
|
|
💼 Deal Execution & Capital (5)
|
🔬 Research & Post-Deal (2)
|
📈 Public Markets (1)
|
See docs/architecture_readme.md for the full architecture documentation with 11 Mermaid diagrams.
Three-pane layout: left nav (sessions, agents, workspace), center chat with streaming responses, right artifact canvas (tables, charts, PDF previews).
- IPO Map — global treemap sized by market cap, colored by post-IPO performance (RdYlGn)
- IPO Pipeline — upcoming US IPOs from NASDAQ calendar, private mega-caps by sector
- SPAC Tracker — KPI cards, annual IPO chart, status donut, searchable table with 13F cross-reference
- Prospectus Builder — analyze IPO prospectus documents
Interactive Plotly treemap of SEC 13F institutional holdings. Filter by fund, min value, and position count. 10K+ funds, 7M+ holdings.
Companies directory, deal pipelines (kanban), daily AI-curated digest, valuation simulator (DCF + multiples), text-to-SQL analytics, virtual data room, document management, deal history.
- Market Comparables — sector M&A benchmarks
- Business Valuation — indicative range with AI value drivers
- Find Buyers — strategic and financial buyer matching
- XLSX — agent tables as formatted spreadsheets
- DOCX — IC memos and teasers as Word documents
- PDF — memo preview in-app, company cards as branded PDFs
uv pip install -r requirements.txt
python main.py # starts on port 5007Or with Docker:
docker compose up --buildXAI_API_KEY=... # primary LLM (Grok via x.ai)
OPENAI_API_KEY=... # fallback LLM
EXA_API_KEY=... # semantic search
TAVILY_API_KEY=... # web search
DB_URL=postgresql://... # PostgreSQL
SESSION_SECRET=... # session cookie signing
At least one of XAI_API_KEY or OPENAI_API_KEY is required.
pytest -q # unit tests (125 cases)
pytest -q tests/test_registry.py # agent registry integrity
pytest -m e2e # Playwright E2E (requires server on :5007)- Backend — Python, FastHTML, HTMX, LangGraph, LangChain
- LLM — Grok (x.ai) or OpenAI via
utils/llm_factory - Database — PostgreSQL (
liquidround+hedgefolioschemas) - Charts — Plotly.js (treemaps, bar, donut, histograms)
- Styling — Tailwind CSS (CDN), custom dark navy + amber palette
- Data — yfinance, SEC EDGAR, Exa, Tavily, NASDAQ calendar
- Deploy — Docker Compose on Coolify, Postmark for email
python -m scripts.sync_spacs --enrich # SPAC data (NASDAQ + EDGAR + yfinance)
python -m scripts.download_sec_13f # quarterly 13F institutional holdings
python -m scripts.sync_activist # daily 13D/G activist filings
python -m scripts.daily_deals --all # send daily digest to opted-in usersProprietary — Predictive Labs Ltd. All rights reserved.
Predictive Labs Ltd · liquidround.ai
