M&A Intelligence MCP server
Let an AI assistant search, summarize, chart and tabulate upstream M&A deals, with 10 tools that act as you.
Ask an assistant such as Claude to find, total, chart and tabulate upstream M&A deals, and keep a comps basket, as you.
Connect
Section titled “Connect”| URL | https:/, on the host you use for Novi apps in the browser |
| Transport | Streamable HTTP |
| Sign-in | Novi sign-in from a supported client, or a Novi API token as Authorization: Bearer |
| Permissions | Access Intelligence M&A. Signing in from an assistant also needs Access MCP |
| Status | Preview Charts and tables (ma_, ma_) are not yet on company hosts |
Add the URL as a custom connector or remote MCP server and sign in with your Novi login. Supported clients and steps: Connect an assistant.
Any client that takes a URL and a header works, with a token for a user with Access Intelligence M&A.
claude mcp add --transport http novi-ma https://<your-company>.novilabs.io/intelligence-ma/api/mcp \ --header "Authorization: Bearer $NOVI_API_TOKEN"If the token expires or is revoked, create a new one and update the client.
| Tool | What it does | Key inputs |
|---|---|---|
ma_ |
Turns a sentence into filters and runs them. Returns the filters, its interpretation, and unsupported and unknown_ for the parts of the question the data can’t express |
query, limit, offset |
ma_ |
Lists deals that match explicit filters. Returns total, limit, offset and transactions |
Filters, sort, dir, limit, offset |
ma_ |
Totals, averages and medians over the matching deals in one call, with breakdowns by deal type, scope, commodity, consideration type and ownership direction | Filters |
ma_ |
Valid filter values as spelled in the data, basin to sub-basin pairings and sort keys. No company names | None |
ma_ |
Full detail for one deal: every column plus basins, states, tags, advisors and the analyst’s comments | deal_ |
ma_ |
Adds deals to your comps basket. Returns the basket’s deal ids, how many were added and unknown_ |
deal_ |
ma_ |
Removes deals from your basket. Returns the basket’s deal ids, how many were removed and not_ |
deal_ |
ma_ |
Your basket in full, oldest first, with the detail ma_ returns. Returns count and transactions |
None |
ma_ |
Charts deals: one bar or point per value of dim, measuring metric |
dim, metric, type, color_, secondary_, top, filters |
ma_ |
Shows matching deals as a table: announced, buyer, seller, type, basins, value, production and price per flowing BOE/d | sort, dir, limit, offset, filters |
Use ma_search_deals in conversation, ma_filter_deals when the filters are decided, and ma_summarize_deals for “how much”.
Filters
Section titled “Filters”ma_filter_deals, ma_summarize_deals, ma_chart_deals and ma_table_deals take these filters. Values in one list are ORed; different filters are ANDed.
| Filter | Takes |
|---|---|
search |
Full text over the deal |
basin, subbasin |
Lists of names |
state, buyer, seller, deal_, deal_, commodity, deal_, other_, consideration_, ownership_ |
Lists of values as ma_ spells them. Buyer and seller names are used as given |
year_, year_ |
Announcement years |
announce_, announce_, close_, close_ |
Dates as YYYY-MM-DD, inclusive. A closing range matches only deals that have closed |
size_, size_ |
Total consideration in $MM |
production_, production_ |
Net production in Mboe/d, so 500 boe/d is 0.5. Matches only deals that disclose production |
liquids_, liquids_ |
Liquids share of production in percent, 0 to 100 |
- Midland, Delaware, Marcellus and Utica are sub-basins. Given as a basin, they match as sub-basins and the result says so. An unknown name is an error that lists the valid names.
- The deal tag
Novi Researchselects deals Novi has published research on. - Liquids filters take a percent; the
LIQUIDScolumn in Snowflake is a fraction from 0 to 1.
Totals and coverage
Section titled “Totals and coverage”Each total says how many deals it covers, since not every deal discloses every figure.
| Figure | Covers |
|---|---|
total_ |
Every matching deal |
total_, avg_, median_ ($MM) |
valued_, the deals that disclose a price |
total_ (Mboe/d) |
production_ |
median_ ($ per BOE/d) |
consideration_ |
Quote the count with the figure, for example “$4,800MM across 12 of 15 deals”.
Charts and tables
Section titled “Charts and tables”In clients that support MCP Apps, such as Claude, ma_chart_deals draws a chart with a table tab and ma_table_deals a sortable table. Every client gets the rows as JSON.
| Setting | Values |
|---|---|
dim (the axis) |
year, close_, deal_, deal_, deal_, commodity, ownership, consideration, buyer, seller, pe, basin, subbasin, geography, deal_, pe_, sponsor_ |
metric |
count (default), value, avg, production, acres, pv10, per_, consideration_ |
type |
bar (default), line for a trend over years, table to open on the table |
top |
How many of the largest to show on an unordered axis such as buyer or basin: 1 to 50, default 20. Years and deal-size bands show in full |
color_ |
Splits each bar into stacked bands by a second dimension. Not with an averaged metric (avg, per_, consideration_) |
secondary_ |
A second measure drawn as a labelled line on a right-hand axis |
On basin, subbasin, geography, deal_tags, pe_firm and sponsor_type, a deal counts under each of its values, so bars can add up to more than the matching deals. The result’s notes say so, and when a figure covers only the deals that disclose it.
Paging, basket and errors
Section titled “Paging, basket and errors”- Paging.
limitis 1 to 100 rows, default 25. Page withoffset. - Comps basket. One per person, kept across sessions, up to 50 deals. Separate from the comps list in the M&A app.
- Deal ids.
deal_idis not theMA_DEAL_KEYin Snowflake. - Errors. An unknown deal id or filter value comes back as a tool result that names the cause.
- Search limits.
ma_search_dealsis rate limited per person; the error says how many seconds to wait.ma_filter_dealsis not affected.
What Novi records
Section titled “What Novi records”For each tool call, Novi records the tool name, how long it took, whether it failed, how many deals it returned, which filters had a value, and the assistant app you signed in from. Filter values and the text of your search are not recorded.
Good prompts
Section titled “Good prompts”- Chart deal count by year since 2015, split by deal size, with total value as a line.
- Show the 15 largest Permian deals since 2020 as a table.
- Chart the top 10 buyers in the Permian since 2020 by total value.
- How much was paid for oil-weighted Williston deals closed since 2023, and how many of them disclosed a price?