Example prompts
Real questions the Dataflow MCP server answers well, and the tools your agent will use for each. Try these verbatim after connecting.
Who owns a company?
Who are the largest owners of Volvo?
The agent resolves the company with search_securities, then pulls the owner list with get_owner_list. Ask for the manager view ("which fund managers hold Volvo?") to switch attribution mode.
Which companies match a profile?
Which Swedish health-care companies are there, and who owns the biggest one?
list_sectors turns "health care" into the sector id (the taxonomy is Modular Finance's own, so it is read rather than guessed), list_companies returns the companies in it ordered by market cap, and get_owner_list takes the company_id straight from the first row. Screens combine sector, industry, country, market segment (list_market_segments for ids such as the FTSE 100) and a market-cap range - "UK-listed industrials above 1bn", "the 20 largest companies on First North".
Who has been buying or selling?
Which pension funds increased their position in Ericsson this quarter?
search_securities resolves the company, then list_owner_changes ranks the period's moves, filterable by change type (new, increased, decreased, exited) and window (up to 1825 days - five years).
What does an investor hold?
What does Alecta's portfolio look like, and what did they change last month?
search_owners resolves the owner, get_owner_portfolio returns every disclosed position, and list_portfolio_changes ranks last month's moves. Add a date, such as "as of June last year", for a historical snapshot (up to five years back).
What large trades printed today?
Show me the biggest block trades on the Swedish market this week.
list_market_block_trades is a market-wide feed - no company needed. countries selects by the market the trade printed on, not where the company is incorporated, so a London print of a Stockholm-primary company is a GB row. Add min_value to keep it to material blocks. Each row names its own company, so the agent can go straight from a print to that company's owners. For one company's block trades, use list_block_trades instead.
Drill into ownership structures
BlackRock holds Atlas Copco. Which of their funds actually hold the position?
Rows with sub-structure carry a has_sub_holdings flag; the agent drills down level by level with get_sub_holdings, from the parent entity to individual funds.
Combine into real workflows
Build a profile of SEB's ownership: the current top owners, who entered or exited in the last year, and how the largest owner's stake has developed.
Agents chain the tools freely: owner lists, changes and portfolios share the same company_id/owner_id identifiers, so multi-step questions work without glue.
Tips
- Be specific about time: "this quarter", "last 90 days", "as of June 2024" map directly onto tool parameters.
- Name the mode when it matters: "beneficial owners" vs "fund managers". See Beneficial vs manager.
- Watch the credits: returned top-level rows are metered; empty results are free. See Credits & pricing.