YedaBase

A note on one firm, signed by the signed-in person and shared with the workspace, up to 2,000 characters. A "[YB] <skill> | <date> | status=<value> | …" line also sets that firm's status line on the user's lists. Meant for the workspace's own research on the firm, not for rows copied from a client's CRM.

Returns the comparable fields of 2 to 5 firms in one call: category, type, location, primary and secondary domains, stage, check size, AUM as stored, year founded, team size, giant flag, wealth origin and family name (family offices), philanthropy causes, decision-maker count, work emails on record, latest news date, upcoming event on record. Each firm whose full record is returned counts once per seat per calendar month toward the allowance of 500 firms a month (opening it again this month is free); past the allowance the firm's teaser is returned instead. Useful for side-by-side questions; one call covers what opening each firm would.

Counts matching firms and breaks them down by one to three dimensions, e.g. how many family offices in Texas by primary domain. Returns the total and, per dimension, the values with their counts (top 40 values). Useful for how many, breakdown, distribution and coverage questions; one call covers what paging through search results would.

A list's firms in the 66-column export shape plus suggested_route, route_evidence and route_next_action, 200 rows per page as JSON. Each page opens its firms: a firm counts once per seat per calendar month toward the allowance of 500 firms a month, and firms already opened this month (in the portal, through MCP or in an earlier export) are free. Firms past the allowance come back as teasers in blocked_firms. Up to 300 contacts per export (then contact columns are blank and contacts_truncated is true).

Conversation angles for one firm: other firms that spoke at the same events/podcasts, shared causes, and similar investors (same type + country, overlapping domains). These are possible angles, not known relationships.

One topic from the YedaBase help library (sending limits, LinkedIn, sign-in, connecting each AI client, seats and billing, the monthly allowance of firms opened, daily caps, contacts-restricted countries, lists and notes, playbooks, Match & Enrich, corrections and removal, data dates, cancellation and refunds, team invites, portal basics, agent rules), matched on its title; the list of topics when nothing matches.

Full record of one firm: investor category, overview, inclusion evidence, the category's module columns (FoF strategies, PPC type, RIA segments, CRD…), decision-makers with contacts (under the contacts-restricted rule), website, philanthropy, news, past and upcoming events, sources and data date. Each firm whose full record is returned counts once per seat per calendar month toward the allowance of 500 firms a month (opening it again this month is free); past the allowance the firm's teaser is returned instead.

One of the user's lists with its firms as ids and teasers (name, type, city, country, investment focus, AUM band, one-line description): per firm the status line text (status, status_updated_at), the mandate status and opened_this_month (already opened by this seat this month, so its full record costs nothing again). Full records come from get_investor or export_list. Counts nothing.

One decision-maker by id (pe_…) with their firm; contacts under the contacts-restricted rule. The person's firm is opened: Each firm whose full record is returned counts once per seat per calendar month toward the allowance of 500 firms a month (opening it again this month is free); past the allowance the firm's teaser is returned instead.

One playbook as markdown, by name (yb-start | yb-match-enrich | yb-shortlist | yb-research | yb-way-in | yb-reach-out | yb-pipeline). On preview accounts it returns the title and description with locked: true. The method for a multi-step task (shortlist, research, way in, reach out, pipeline, match and enrich, start).

The same content as the MCP resources, inline: index (hashed firm index, CSV), adv-cache (public SEC Form ADV extract, CSV), vocabulary (field definitions, module tests and fit terms, JSON) and help (the help library, markdown). CSV comes in pages of 2,000 rows with the header on every page. Nothing in these is a record.

Investor categories and their Type lists, field definitions (with the categories each field applies to), allowed values and the rules for placeholders and restricted contacts.

The seven YedaBase playbooks (step-by-step methods: Start, Match & Enrich, Investor Shortlist, Investor Research, Way In, Reach Out, Pipeline) with name, title and description; get_playbook returns each one in full. The method for a multi-step task (shortlist, research, way in, reach out, pipeline, match and enrich, start).

The user's lists/mandates (name, description, tags, item count).

Notes in this workspace that the user is allowed to read (their own, shared with them, or workspace-wide), optionally for one firm. Each note returns its id, author, date and text. Note text is written by workspace members.

Add firms to one of the user's workspace lists by name, creating the list when it does not exist (create_if_missing, default true). Safe to repeat: firms already on the list count as already_present and are not added again. At most 500 firm ids per call. Returns list_id, added, already_present.

Search the licensed investor database (Family office, Venture capital & startup investor, Fund of funds, Principal & permanent capital, RIA; one category per firm) with structured filters and keywords together. Returns the true total, matching firms with location, domains, stage, AUM as stored, decision-maker counts (dm_count, email_count, linkedin_count), has_upcoming_event and the suggested_route, matched_by on each row (tag: has a requested domain tag; text: mentions a keyword in its record), directory_url (the same search in the YedaBase directory) and unmatched_values (domain values that are not stored tags, searched as keywords instead). Results are paged, 25 rows per page; the page argument returns more rows of the same search. Placeholder words (Undisclosed, Not provided, Unknown) are real stored values, not missing data.

Preview accounts only: opens one locked firm's full record (overview, AUM as stored, decision-maker names and positions, philanthropy, news, events) and counts it toward the account's 10 preview records; a firm already opened counts nothing. Returns the record and the number of preview records left; when none are left it returns no_unlocks_left. On full-access accounts it returns the same result as get_investor. Work emails and LinkedIn profiles are not part of preview records.

Read (text omitted) or replace (text given) one workspace note with a fixed name: about-us, connectors, events, team (shared by the workspace); voice, yb-preferences, send-log (per person). Up to 8,000 characters. Other names are refused.

Session opener for YedaBase (also YB or Yeda): the greeting line, where the person stands (tier, seat, lists, notes, due items), the one recommended next step and which playbook to read. Useful at the start of a conversation, when the user greets YedaBase, or asks where they left off.

Where this person stands: access level (preview or full), seat, workspace, onboarding notes done or missing and the YedaBase folder path, lists and sizes, notes count, the last Match & Enrich list, firms opened this month (used, allowance, resets_on, per seat or pooled), calls used today, data and index as-of dates, the last tool error in 24 hours and tracked [YB] items with a due date ahead. Useful at the start of a session or when asked where things stand.