multiples_compute_deal_stats
ChatGPTCompute distributional stats (count, p25, median, p75, mean, min, max) over a filtered set of deals. USE THIS for any "median / typical / average / p25 / p75 X" question about deals - NEVER compute medians from a list_tool sample, the answer will be wrong. - metric: "ev_revenue" / "ev_ebitda" (multiples) | "valuation" (raw post-money, USD M) | "raised" (round size, USD M). - deal_subtypes: filter by round stage (Series A, Seed, Strategic M&A, PE buyout, …) - exact match. - participant_names: filter to deals where a named buyer/investor participated ("Salesforce acquisitions", "rounds with a16z"). Use lead_only=true for "X-led" rounds. - participant_verticals / participant_themes / participant_geographies: filter by the buyer/investor company's own profile ("deals bought by European PE firms"). - bucket_by: "year" or "quarter" for evolution/trend questions ("how have Series A valuations moved", "SaaS M&A multiples 2022 vs 2024"). Server caps at the last 12 buckets - always pair with a sensible deal_date_from when bucketing. Returns one row per bucket (or one row total when not bucketed). NEVER returns a list of individual deals.
multiples_get_time_series_aggregate
ChatGPTCross-sectional p25 / median / p75 / count for one metric across a peer group of 3-50 public companies, computed at each qualifying month then optionally thinned. The 'resolution' input controls density: - 'auto' (default): up to 12 evenly-spaced dated rows. Cheap. Use for first-pass answers. - 'monthly': every qualifying month within the window, no gaps. Use ONLY when the user explicitly asks for month-by-month / no skips / the full series. The result envelope tells you which mode ran (series_mode) and how to escalate (escalation_hint). Never invent explanations like "reporting periods" for sampled gaps - escalate to 'monthly' instead. For a single company's trajectory use get_time_series_stats; for one date use get_time_series_snapshot.
multiples_get_time_series_snapshot
ChatGPTRead up to 6 metrics for up to 25 companies at one specific date. Use for "what was Microsoft's EV/Revenue in Q4 2023", "cyber multiples as of Dec 2022 vs today" (two calls), "how did Snowflake look at IPO". For trajectory questions use get_time_series_stats; for peer-group medians over time use get_time_series_aggregate.
multiples_get_time_series_stats
ChatGPTTrajectory of one metric for specific public companies over a date window. Returns per-company scalars (start, latest, change %, min, max, median, observation count) plus a 'series' array. The 'resolution' input controls the series: - 'auto' (default): up to 12 evenly-spaced sampled points per company. Cheap. Use for first-pass answers where prose + scalars suffice. - 'monthly': every month within the window, no gaps. Use ONLY when the user explicitly asks for month-by-month / every month / no skips / the full series. Max 5 companies in this mode. The result envelope tells you which mode ran (series_mode) and how to escalate (escalation_hint). Never invent explanations like "reporting periods" for sampled gaps - escalate to 'monthly' instead. For peer-group medians use get_time_series_aggregate; for a single date use get_time_series_snapshot.
multiples_list_taxonomy_values
ChatGPTReturns canonical filter values across taxonomies (verticals/themes/geographies/deal_subtypes/client_focus/revenue_model/company_type). Set filters in search tools are EXACT MATCH against these values. Returns hierarchical structure: rows have {taxonomy, value, kind: parent|leaf, parent?, children?}. When kind=parent, the underlying data is tagged at leaf level - pass parent + children together as filter values to match all child-tagged rows. OMIT type for ambiguous concepts to search across all taxonomies at once.
multiples_search_buyer_rankings
ChatGPTRanks M&A acquirers by deal count over the FULL filtered set. USE THIS for "most acquisitive buyers in <X>", "top strategic acquirers", "most active PE buyers", "who has bought the most <X> companies" questions. ALWAYS set deal_subtypes when the user implies a buyer type: "strategic" / "corporates" → ["Strategic M&A"]; "PE" / "financial sponsors" / "sponsors" → ["PE buyout"]. Omitting deal_subtypes mixes strategics and PE funds together and is rarely what the user wants. Returns one row per buyer with deal_count, total disclosed transaction value, latest deal date, and buyer profile fields.
multiples_search_companies_by_name
ChatGPTLook up companies by name (ilike contains). Returns id, name, short_name, plus listed_in_public_dataset + public_company_name + ticker + exchange when the company is in the public-comps dataset. ALWAYS call this FIRST for "tell me about <company>" or "info on <company>" questions - if listed_in_public_dataset is true, follow up with search_public_companies passing companies:[id] (the id from this result) to pull live multiples. Pass the id, not the name: names are not unique (two listed companies are both called "Zoom") and a name pulls every match.
multiples_search_funding_rounds
ChatGPTSearch funding rounds (VC, growth, IPO, secondary - anything that isn't M&A). Numeric monetary inputs are USD millions. Returns curated rows: deal_id, deal_date, target_name, deal_subtype, deal_raised, deal_valuation, EV multiples, investors, lead_investors, target country/verticals.
multiples_search_investor_rankings
ChatGPTRanks investors by funding-round participation count over the FULL filtered set. USE THIS for "most active VCs in X", "top investors in X", "biggest backers of X", "who has invested the most in X", "top lead investors in X" - any question that asks to rank investors. Set lead_only=true ONLY when the user explicitly says "lead investors" / "rounds led by" / "who led". Returns one row per investor with deal_count, lead_count (how many were leads), distinct_companies, total disclosed raise, latest deal date.
multiples_search_ma_deals
ChatGPTSearch M&A transactions. Use deal_subtypes:["Strategic M&A"] for strategic / corporate acquisitions, deal_subtypes:["PE buyout"] for PE buyouts / LBOs. Filters AND together. Numeric monetary inputs are USD millions. Returns curated rows: deal_id, deal_date, target/buyer names + countries + verticals, deal_valuation/revenue/ebitda, EV multiples. The buyers array lists ALL acquirers with id, name, country (ISO code), and current ownership. Use it when identifying buyers; buyer_name and other scalar buyer fields describe only the primary buyer.
multiples_search_public_companies
ChatGPTSearch listed-company benchmarks (public_companies). Numeric monetary inputs are USD millions. Returns curated rows: company_name, ticker, exchange, country_name, verticals, market_cap, enterprise_value, FY actuals (fy_year, revenue_fy, ebitda_fy, gross_profit_fy, net_income_fy, fcf_fy), EV multiples (LTM and NTM), revenue growth NTM/LTM, EBITDA margin LTM. Raw LTM and NTM P&L figures (revenue / EBITDA / EBIT / net income / FCF / gross profit) are intentionally NOT in the returned rows - use the FY actuals for "what is X's revenue / gross profit / etc." questions. revenue_min/max and ebitda_min/max still filter on the chosen period (LTM default) and sort_field still accepts revenue_ltm - filtering and ranking work, the underlying value is just not disclosed.
multiples_search_sector_rankings
ChatGPTRanks sectors (verticals, themes, or countries) by deal count over the FULL filtered set. USE THIS for "hottest verticals / themes / geographies in <X>", "where is the most M&A / funding activity", "most active sectors right now". For "deals" by company use search_top_companies; for individual sector deep-dives use compute_deal_stats. Returns one row per sector with deal_count, total disclosed valuation, median valuation, latest deal date.
multiples_search_top_companies
ChatGPTReturns companies ranked by their maximum known valuation across all funding rounds and M&A deals (deduplicated by company). USE THIS for "largest <X>", "biggest <X>", "top private companies", "leading neobanks" type questions. Aggregates over the FULL filtered dataset server-side; you do not need to dedupe or rank in your head.
multiples_compute_deal_stats
Claudemultiples_get_time_series_aggregate
Claudemultiples_get_time_series_snapshot
Claudemultiples_get_time_series_stats
Claudemultiples_list_taxonomy_values
Claudemultiples_search_buyer_rankings
Claudemultiples_search_companies_by_name
Claudemultiples_search_funding_rounds
Claudemultiples_search_investor_rankings
Claudemultiples_search_ma_deals
Claudemultiples_search_public_companies
Claudemultiples_search_sector_rankings
Claudemultiples_search_top_companies
Claude
