MCP Tools for Event-Driven Trading Research

Use Alphanume's event, regulatory, and point-in-time reference tools to build dated cohorts that can be audited before they become backtests.

Event-driven research starts with a timestamped event, not a list of stocks that look unusual today. Alphanume's MCP server exposes separate tools for dilution filings, shelf registrations, completed de-SPAC transactions, corporate defaults, SEC trading suspensions, and reference data. An agent can query those tools conversationally, but the useful output is still a cohort with explicit dates, fields, and inclusion rules.

The MCP layer returns the same deterministic point-in-time records as the Alphanume REST API. It reduces query plumbing and helps inspect schemas, but it does not decide whether an event is causal, whether a sample is large enough, or whether a strategy survives spreads, borrow, and execution costs. Those decisions remain part of the research design.

Map the event tools to distinct hypotheses

The Dilution, Shelf Registrations, De-SPAC Events, Corporate Defaults, SEC Trading Suspensions, and Historical Market Cap references define the fields behind these tools. Use the MCP guide for authentication and response conventions.

Build a small-cap dilution cohort

A concrete first study is a cohort of S-1 dilution filings from smaller issuers. Query get_dilution_filings over a fixed filing-date range. Retain the filing timestamp, ticker, market cap at filing, dilutive and resale flags, shares offered, effectiveness state, withdrawal state, accession number, and filing URL. Then check the ticker's point-in-time capitalization with get_historical_market_cap on the same information date.

Use only Alphanume MCP tools.

Build a small-cap dilution research cohort for one date range available to my account.

1. Call get_dilution_filings with date_gte and date_lte.
2. Keep the filing timestamp, ticker, market cap at filing, dilutive flag, resale flag, shares offered, effectiveness fields, withdrawal fields, accession number, and filing URL.
3. For each ticker, call get_historical_market_cap for the event information date.
4. Keep rows only when a point-in-time market-cap observation is available and apply a size threshold that I supply separately.
5. Return excluded rows in a second table with the exclusion reason.
6. State every tool, filter, row count, null field, and tier restriction.
7. Do not calculate post-event performance or call the cohort a strategy.

The prompt deliberately leaves the size threshold to the researcher. A model should not invent one and then present the resulting cohort as inevitable. It also requires an exclusions table, which exposes missing market-cap matches, null tickers, and records outside the account window instead of allowing silent sample shrinkage.

Separate authorization from issuance

Shelf registrations need their own cohort logic. A shelf can register future issuance capacity, including primary, resale, amendment, and automatic shelf structures. capacity_amount can be null for an indeterminate or automatic shelf and must not be guessed. The presence of registered capacity does not mean the company sold that amount. Takedown counts, EFFECT dates, and later filings are separate lifecycle evidence.

The older S-1 dilution tracker and the shelf-registration ledger overlap in theme but not in contract. Mixing them into a single "dilution happened" flag discards the distinction between a registration, an effective registration, a selling-stockholder resale, and completed issuance. Preserve form, shelf type, resale status, lifecycle dates, and source filing identifiers before defining any combined signal.

Add event families without flattening them

De-SPAC completions, defaults, and trading suspensions should enter as separate event families. A de-SPAC record identifies a completed business combination and may include redemption, PIPE, proceeds, and listing fields. A corporate-default record describes an obligation and may include principal, missed payment, grace-period, and acceleration details. A suspension record describes an SEC order and the legal suspension window. None of those fields are interchangeable.

• Use one anchor-date definition per family and keep it in the output.
• Retain the source filing or order URL so a surprising row can be inspected.
• Keep null enriched fields as null rather than treating missing disclosure as zero.
• Normalize outcomes only after the individual event tables pass field-level checks.
• Use current ticker classification only when a current-state sector label is acceptable for the study.

Audit timing, pagination, and access

Every Alphanume data tool accepts max_rows, which defaults to 500 at the MCP layer. Wide tools can return pagination metadata, and Historical Market Cap requires both cursor date and cursor ticker to continue a large response. A cohort that stops at the first page is a truncated sample. A cohort that uses today's market cap for an old filing contains lookahead. A cohort that drops null enriched fields can become a disclosure-quality screen without saying so.

Free access provides the trailing 20 trading sessions after a one-trading-session delay. Pro provides current data and full history. Some event types are sparse, so an empty recent window can be valid. Distinguish an empty response from a tier restriction, authentication failure, or row cap before changing the cohort.

Audit the cohort you just built.

Report the requested date range, account restrictions encountered, tool-level row counts, pagination state, excluded-row counts by reason, duplicate event identifiers, null rates for every retained field, and the earliest information date used in each join. Stop if any page is missing or if a reference observation comes from after its event date.

Save a research-ready cohort

Export the final event table together with its audit output, tool names, filters, source URLs, and an as-of timestamp. That bundle is the input to an event study, not the conclusion of one. Define the return window, benchmark, delisting treatment, transaction costs, and out-of-sample period separately. For a fuller treatment of the research process, use Systematic Event-Driven Trading as related methodology, then connect through the hosted MCP server and run the narrow cohort prompt before expanding the history on the pricing page.