MCP App Store
by KeenEthics
Monk.trade icon

Monk.trade MCP Server for ChatGPT

By LabStack LLC

Overview

Monk lets your AI scan US stocks, ETFs and futures, chart the results, and cite supporting evidence when a scan reads SEC filings, insider trades or news, through MonkQL, Monk's market query language. Your AI writes MonkQL to quote and chart symbols, scan the market with your own conditions, and keep private saved queries with their full version history. Results show their values and available evidence. Monk never trades or reaches a brokerage.

How to connect

  1. ChatGPT: Open the Monk.trade listing in the ChatGPT app directory and choose Connect. ChatGPT listing Monk.trade exposes 7 tools here.

Tools

create_query

ChatGPT
Saves a private named query or batch. Text must compile; saving pins every use line to its library digest and records version 1. Names are unique per owner, case-insensitively. Identical input with the same idempotency_key replays the receipt within 24 hours, including after uncertain outcomes. Changed input with the same key fails.

delete_query

ChatGPT
Permanently deletes an owned saved query and all its version history by UUID. This cannot be undone. A repeated delete returns not-found.

list_queries

ChatGPT
List your saved queries, newest updated first. Takes no arguments.

list_query_versions

ChatGPT
Lists append-only versions of an owned saved query by UUID, newest first, with version, name, description, text, digest and created_at. run_saved_query with version runs old text; update_query with old text and the current version as expected_version creates a new version restoring it.

run_query

ChatGPT
MonkQL: US stock/futures quotes, charts, scans, filings, insider trades, news, fundamentals. Research only; no orders. Defaults: stocks/all, daily regular. Clause order is free; fn/let keep their order. from names quoted symbols or universe <slug>; from stocks/futures alone selects that market’s universe all. Scans need where or sort; named symbols need neither. include sets rank/percentile population; where filters afterward. sort is numeric asc/desc; take defaults 100, max 500. Show names are canonical expressions (48 chars, numbered duplicates); sort by expression. avg/mean(x, n) = sma(x, n); sum(condition, n) = count(condition, n); if(c, a, b) = c ? a : b. x[n] is n bars ago, never x@n or lag(); x@1h reads an equal/larger interval. Quote: from "IBM" show close, close[1] as prior Rank: include type == "common" where rank(roc(close, 20) by sector) == 1 show sector, industry, market_cap, shares_outstanding, pe, ps, pb, percentile(roc(close, 20)) as pct sort pct desc Labels: type includes "common", "etf", "adr"; sector/industry/name are labels; futures category = sector (Metals); name never groups. industry_group/sub_industry are unpopulated. Rank/percentile groups <20 are undecided; warnings count groups/symbols; all groups small fails; rank 1 is highest, percentile is 0-100. Premarket: on 5m premarket show 100 (close / close@daily - 1) as gap, volume@session as vol sort gap desc Afterhours: on 5m afterhours show 100 (close / close@regular - 1) as move, high@premarket as prehigh sort move desc Phases: 5m, 15m, 30m, 1h, 4h; on premarket/regular/afterhours defaults to 5m; 1d = daily, no daily phase or 1m. @session: phase volume/price so far; @premarket/@regular: that phase; close@daily: prior session's close before open, today's after; premarket gap = close / close@daily - 1; close@daily[1] is two sessions back. Notional = close volume. Bare vwap: cumulative ((high + low + close) / 3) volume / cumulative volume, reset each New York phase; intraday stocks only (on 5m), stored venue-subset volume, slightly different from consolidated VWAP. Insiders: where insiders(code = "p", relation = "officer", security = "common", plan = "none", window = 20) window/within/lookback = n reads n query bars; canonical window = n: condition = count(dataset, n) > 0; .count totals counts; total fields sum admitted values, e.g. insiders(window = 63).net. Newest fields read last(field, n); novelty/hourly and facts windows are errors. News: where news(topic = "earnings", window = 5) Facts: where facts(concept = "revenue", period = "ttm").yoy > 0.10 show facts(concept = "revenue").value as revenue .value is reported; .yoy is a ratio (0.10 = 10%); period "quarter" = "quarterly", "year"/"yearly" = "annual", "ttm" = trailing year. Filings: where filings(form = "13d", kind = "amendment", change = "added") show filings(form = "13d").holder as holder, filings(form = "13d").stake as stake sort stake desc Forms: 8-k, 10-k, 10-q, 20-f, 40-f, 6-k, nt 10-k, nt 10-q, 10-12b, s-1, s-3, 424b4, 424b5, 13d, 13g, sc to-t, sc to-i, sc to-c, sc 14d9. kind: original/amendment/preliminary; change: new/added/reduced/unchanged/exited/switched/unknown. 8-K items: filings(form = "8-k", item = "2.02"); count(filings(form = "13g"), 5) counts filings. Fields: count, prior, lag, lag_before, holder, stake (percent), stake_before, stake_change (points), shares. Filters include role, session, holder, text. Expressions: from "IBM" fn body(k) = abs(close[k] - open[k]) show (close > open ? body(0) : na) ?? 0 as body_size, sum(i in 0..4 : body(i)) as total_body Library: use "<id>" as lib; lib.f(x). Other symbol: close of "SPY". Chart: from "IBM" on 5m mark level close[1] label "Prior" plot ema(close, 20)@1h chart Plots/marks accept comma lists; plot x as name labels it. Both need chart (max five named symbols): newest 200 bars, overlays and optional PNGs. A line --- batches five independent queries in 8192 UTF-8 bytes. Stock all daily keeps active stocks (at least $20M of close × volume in total over the last 20 trading sessions (about $1M a day), or $1M on the newest sessio…

run_saved_query

ChatGPT
Runs an owned saved query by UUID, optionally at a historical version (integer at least 1). Omitted version selects current text. Results use the same public contract as run_query, including run_id, status, data_status, counts, warnings and undecided symbol/reason pairs. A result with a problem or diagnostics did not run as written: correct the query from them and run it again. Charts use detail:full by default with the newest 200 available bars, aligned plots and marks. Hosts that render Monk's view receive detail:compact with the last 20 bars and a summary, with the full chart in the view. MCP also attaches PNGs when available. chart full is equivalent to chart; query.saved reports the id and version of the text that ran. list_query_versions reads append-only history; update_query with old text and the current version as expected_version creates a new version restoring that text.

update_query

ChatGPT
Changes only supplied name, text or description fields of an owned saved query. expected_version is the version the change is based on; a stale or missing value changes nothing and returns version-conflict with the current query. At least one field is required. Values unchanged after library pinning return the current version unchanged; a changed field appends a new version. Omitted fields stay unchanged; an empty description clears it. Supplying an older version's text creates a new version restoring it.

Capabilities

InteractiveWrites

Example Prompts

Click any prompt to copy it.

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/monk-trade"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/monk-trade)