SPAC data, the way it should have been built.

The active SPAC universe — currently 1,298 entities spanning every lifecycle state (searching, in vote, combination announced, closed, liquidated). For each, we maintain identifiers (CIK, CUSIP, ticker), trust economics, sponsor + advisor graph, deadline + extension history, votes + redemptions, and downstream deSPAC outcomes.

Trust values are anchored on quarterly iXBRL observations from 10-Q / 10-K filings and interpolated daily using FRED 3-month T-bill rates. Implicit redemptions are inferred when the trust balance steps down between quarters without a vote — a signal most data providers miss.

• Citation-first. Every extracted number deep-links to the source SEC filing in our in-app viewer with the cited value highlighted on the page. Click any 0001234567-25-001234 chip and you land on the exact line that backs the data.
• Real-time agentic alerts. Eight rule kinds fire on extension votes, ≥75% redemptions, ≥40% trust drawdowns, lockup expiries, deSPAC closings, IPO launches, liquidations, and YTM-arb opportunities. Stream live via /api/v1/alerts/stream.
• Conversational query. /ask converts plain English to SQL against the live database. "SPACs with extension votes in the last 90 days where redemption was over 50% of trust" returns a clean tabular answer in under five seconds.
• API-first. JSON over HTTP on every dataset (universe, calendar, alerts, news, search, individual SPACs). API keys gate quota; bearer auth via Authorization: Bearer sk_live_….
• Editorial layer. A daily SpacDesk Newsroom digest stitched from yesterday's structured events; monthly lockup watchlists projected from charter conventions.

Our pipeline is open by design. EDGAR filings flow through a three-stage extractor: (1) iXBRL companyfacts where structured tags exist, (2) heuristic regex over standard 8-K / S-1 / DEF 14A templates, (3) LLM long-tail extraction (Claude Sonnet 4.6 + Haiku 4.5 verifier with strict JSON schema) for the messy remainder. A 48-row pinned XBRL eval enforces 100% pass-rate as a regression guard.

We're built on Postgres 16, Next.js 16, and Drizzle ORM. The ingest sidecar is Python 3.11 + edgartools. Hosted at spacdesk.com via Cloudflare Tunnel.

SpacDesk is in early access. The core data product is live and updated continuously by background crons (filings poll, news RSS, XBRL backfill, alerts evaluator). Authentication and billing are scaffolded and will activate as we onboard paying customers.

Have feedback, want a demo, or want to integrate? Email [email protected].