Massive

Fetch data from a Massive.com REST API endpoint. Use a path from search_endpoints results. Set store_as to save results as an in-memory table for SQL querying with query_data; the first store auto-creates a workspace and returns its handle — pass that workspace= to query_data and to later stores to keep them in one scope. Paginated responses include a next-page hint with the exact path and params for the follow-up request. The apply parameter runs built-in functions on results — string input values refer to table columns, numeric values are literals. Use search_endpoints with scope="functions" to discover available functions.

call_api

Manage isolated scratch workspaces for storing API results. A workspace handle is the per-conversation scope for stored tables. A handle is created automatically the first time you call call_api with store_as (it's returned in the result) — use this tool only to create one up front, list existing ones, or delete one. One workspace = one isolated scratch space; create separate handles for work that must not share data, reuse one (e.g. across sub-agents) when sharing is intended. Workspaces expire after 1h idle. To resume or switch, use action='list' and pass the chosen handle onward — there is no server-side "active" workspace. Use action='delete' to free one early. If a call reports the handle is unknown/expired, create a new one and re-stage data — don't retry the dead handle.

workspace

Run SQL queries on tables stored via call_api's store_as parameter. Special commands: SHOW TABLES, DESCRIBE <table>, DROP TABLE <table>. Tables auto-expire after 1 hour. Supports CTEs, window functions, JOINs, and ILIKE. Queries run in the SQLite dialect. SHOW TABLES lists every stored table with its column names (table_name,rows,columns,bytes,age_seconds,column_names), so it is usually enough to write correct SQL; run DESCRIBE <table> for the column types, or when a wide table's column_names ends in "(+K more)". A "no such column" error also lists the table's actual columns. Full-text search: any stored table with TEXT columns is searchable directly via FTS5. Use `WHERE {table} MATCH 'query'` and ORDER BY rank (lower = better). Numeric columns preserve their types. For long TEXT fields (news body, 10-K risk factors, filings) prefer snippet()/highlight() in SELECT. Recommended FTS pattern: 1. Find matches compactly: `SELECT rowid, category, bm25({t}) AS score, snippet({t}, 4, '[', ']', '...', 15) AS snip FROM {t} WHERE {t} MATCH 'supply chain OR supplier' ORDER BY rank` 2. Fetch the full body only for rows you need: call query_data again with `SELECT supporting_text FROM {t} WHERE rowid IN (2, 7)` and `max_cell_chars=0`. Output cells over max_cell_chars (default 2000) are truncated with a visible marker; raise or set to 0 to disable.

query_data

Search for market data API endpoints and built-in finance functions by natural language query, or fetch full documentation for one endpoint by passing its exact path. Use this FIRST to find the right endpoint before calling call_api. Covers stocks, options, forex, crypto, futures, indices, ETFs, and economic data. Search results include each endpoint's parameter names (required ones marked *) — usually enough to call it directly; pass path="..." exactly as shown in results to get full parameter docs for that one endpoint, adding detail="verbose" if you need to know what the response fields mean. Use detail="more" only when you need parameter docs for several candidates at once. Pass market to pin results to a specific asset class when you already know it; omit it and the server will infer from the query.

search_endpoints