Public deal supply
Reads published broker teaser pages, preserves the source payload, and records each crawl so a quiet or broken source is visible.
MCPs / DealSift MCP
Deal supply arrives as a hundred broker emails and the same business listed three times under three names. DealSift MCP gathers the public listings, resolves the duplicates into one canonical deal, searches the corpus by meaning, and checks it against the thesis the firm actually buys on. What comes back carries its source listing, so an associate can verify a match instead of re-finding it.
The operating loop
The repetitive supply work — the crawling, the deduplication, the watching for a price cut — happens underneath. What reaches a person is evidence with the source attached, at the point where judgment is the only thing left to add.
Reads published broker teaser pages, preserves the source payload, and records each crawl so a quiet or broken source is visible.
Joins cross-posted listings only when title similarity and every shared financial figure agree. The cheaper error is a redundant row, not a false merge.
Searches listing descriptions semantically, then lets the operator narrow by sector, geography, source, status, asking price, EBITDA, revenue, and freshness.
Keeps first-seen and last-seen dates and records discoveries, price changes, relistings, and delistings instead of treating every listing as a static snapshot.
The buy box
Configure the sectors, EBITDA range, and maximum asking price once, and the screen that usually happens in somebody’s head runs against every new listing. Each flag arrives with the reasons attached, so the first question in the meeting is whether to call the broker.
Why DealSift flagged this
Agent access
Connect Claude, ChatGPT, Claude Code, or another approved agent to the authenticated MCP endpoint. The same source-backed records answer the question whether it starts in chat, code, or the DealSift interface. The corpus is built from published broker teasers, it is read-only, and the endpoint reaches nothing else the firm owns.
search_deals
Returns meaning-ranked candidates for the question at hand; structured filters can narrow the same corpus.
get_deal
Returns the description, financials, location, broker listing, live source URL, and observed history for one deal.
Operator view
Explore exposes semantic search, structured filters, thesis flags, favorites, and deal detail. Thesis shows the standing buy box and the deals currently matching it. Both sit on the same corpus the MCP uses.
DealSift MCP
DealSift supplies the live evidence. A Busy teammate owns the mandate, the judgment and the follow-through — and every approach to a broker is drafted for a partner to read and send.