mrmarket.ai

ai.mrmarketv1.1.0更新于 Sep 29, 2026

Analytical engine for US-listed equities: screens, rankings, and event studies in plain English.

已验证Streamable HTTP可网页运行AI & ML

安装

在 SourceWeft 中

  1. 打开 控制台中的 mrmarket.ai，将其添加到工作区。
  2. 为需要使用其工具的对话启用该服务。

Web executable，通过 Streamable HTTP。 远程服务在工作区中配置后即可从网页运行时运行。

其他 MCP 客户端

把它添加到你客户端的 mcpServers 配置中。

{
  "mcpServers": {
    "mrmarket-mcp": {
      "type": "http",
      "url": "https://mcp.mrmarket.ai/mcp"
    }
  }
}

README

mrmarket.ai MCP server

A purpose-built analytical engine for financial research, exposed over MCP. Ask hard questions about US-listed equities in plain English and get structured, reproducible results back.

mrmarket.ai is a remote MCP server. Connect it to Claude, ChatGPT, or Gemini and your assistant gains a dedicated engine for resolving financial questions against clean, structured market data. It is not a language model with a database bolted on, and it is not an API wrapper. You describe the analysis; the engine computes it and returns rows, columns, and the assumptions it applied.

The interesting questions in equity research require joins that no single API endpoint will run for you. "Stocks where insiders bought over $1M, free cash flow grew four quarters running, and the price sits below the 200-day moving average" spans insider filings, quarterly cash flow statements, and daily prices at once. mrmarket.ai is built for exactly that class of question.

Why it is different

  • Computed, not predicted. Answers are derived from the underlying data, not generated from a model's recollection. The same question returns the same answer every time.
  • Cross-dataset by design. Fundamentals, prices, earnings, and insider activity are resolved together in a single question, with time dimensions aligned across differing data frequencies.
  • Point-in-time aware. Historical and as-of questions respect the information available on the date in question, so backward-looking studies do not leak future data.
  • Built to fail loud. When a question falls outside coverage, the engine returns an explicit error rather than fabricating a plausible number.
  • Transparent. Every result carries the assumptions and defaults that were applied, so they can be inspected and adjusted.

Connect

mrmarket.ai is a remote MCP server reachable at:

https://mcp.mrmarket.ai/mcp

Each connection is tied to a free account so that credit balance and rate limits apply correctly. Authorization happens over OAuth on first use. Create an account and manage connections at mrmarket.ai/connect.

Claude Code

bash
claude mcp add --transport http mrmarket https://mcp.mrmarket.ai/mcp

Claude Desktop

Open Settings, go to Connectors, choose Add custom connector, and paste the endpoint:

https://mcp.mrmarket.ai/mcp

Claude prompts for authorization on first use. Note that editing claude_desktop_config.json by hand does not work for this server; that file only accepts local, command-based servers, and mrmarket.ai is remote.

ChatGPT

In Settings, open Connectors, choose Add custom connector, and point it at the same endpoint.

Gemini

Add it as a remote MCP extension when prompted by Gemini's MCP onboarding, using the same endpoint.

Tools

ToolCostPurpose
query_datameteredThe workhorse. A natural-language financial question in; structured rows and columns out. Screens, rankings, comparisons, time-series, cohort studies, and event studies are all single calls.
fetch_pagefreeThe continuation. When a query_data result is too large for one response, page 1 comes back with a pagination.next_cursor; pass that cursor to fetch_page to retrieve the rest. It runs no new query, so it never costs credits — the original query_data is charged once.
describe_datafreeThe data catalog. Browse categories and fields, or search for a specific metric, before composing a question.
get_symbolsfreeFast ticker, sector, and industry resolution. Name to ticker, sector membership, or the full universe in under 50ms.
getting_startedfreeA structured orientation tour with verified example prompts, grouped by use case.
recent_queriesfreeYour recent queries, for context and re-runs.
get_account_statusfreeConnected account, plan tier, credit balance, and rate limits. A useful first call in any new session.
report_issuefreeFlag a result that looks wrong or a feature that is missing. Pass the query_id from the response so the trace can be investigated.

Seven of the eight tools cost zero credits. Only query_data consumes credits, priced by query complexity. Start with getting_started or describe_data, then move to query_data.

What you can ask

Every prompt below resolves in a single query_data call. Phrase questions the way an analyst would speak them, not as a list of column filters.

Screens

Technology stocks with ROIC above 20%, debt-to-equity below 0.5, and 5-year revenue CAGR above 10%.Healthcare companies with positive free cash flow and gross margin above 50%.Companies with four consecutive quarters of growing free cash flow.

Rankings

Top 20 stocks by ROIC, excluding financials.Top 15 companies by smoothness of year-over-year revenue growth over the last five fiscal years.Bottom 10 consumer discretionary stocks by 1-year total return.

Comparisons

Compare AAPL, MSFT, and GOOGL on revenue and net income over the last five years.AMZN versus WMT operating margin trend over the last decade.NVDA quarterly EPS surprise history.

Cohort-relative analysis

Stocks whose ROIC is at least one standard deviation above their sector average.Companies with a net profit margin more than twice their sector average.Companies with revenue growth in the top quartile of their industry.

Time-series and overlays

AAPL daily closes for the last five years with earnings dates overlaid.MSFT price with insider buy and sell markers over the past two years.50-day and 200-day moving averages for TSLA.

Event studies

Average 30-day return after companies beat earnings by more than 10 percent.Forward 3-month returns following insider purchases above $1M, grouped by sector.Across the large-cap universe, the correlation between monthly insider purchase count and the next 3-month return.

Multi-step research, driven by your assistant

Larger studies decompose into several calls that your assistant orchestrates while mrmarket.ai supplies the ground truth at each step:

For every large-cap stock that gapped down 5 percent or more after earnings in thepast two years, what is the 30, 60, and 90-day recovery rate, and is it statisticallysignificant against random 90-day windows?
Backtest this signal: buy stocks with insider purchases above $500K within 30 daysbefore earnings that then beat EPS estimates, hold for 63 trading days, and reportCAGR, maximum drawdown, and win rate against buying every post-beat stock.
Build a pre-earnings watchlist: companies reporting in the next two weeks whereinsiders bought in the last 90 days, the stock trades below its 200-day movingaverage, and they beat estimates last quarter.

Coverage

Universe11,000+ US-listed stocks and ETFs (NYSE, NASDAQ, AMEX)
History20+ years of financials on US equities
Response timeUnder 10 seconds for typical questions

Data spans daily adjusted prices, annual and quarterly financial statements (income statement, balance sheet, cash flow), earnings actuals against estimates, insider transactions, and valuation snapshots, alongside computed metrics such as ROIC, free cash flow, debt-to-equity, ROE, margins, returns, and earnings surprise. Sourced with point-in-time accuracy and curated for cross-dataset queries.

Scope

mrmarket.ai is a research and analysis layer, not a brokerage and not a real-time feed. Prices are end-of-day. The following are out of scope: intraday and tick data, options chains, news and transcript text, raw filing text, macro series, FX, crypto, bonds, and futures. When a question reaches an edge, the engine reports it rather than guessing.

Troubleshooting

Every query_data response carries a stable error_code, plus warnings and assumptions arrays that explain how the question was interpreted. Read those first; they usually point at the cause.

SymptomWhat it meansWhat to do
QUERY_NOT_UNDERSTOODThe question was ambiguous or referenced an unsupported metric.Rephrase with specifics (period type, time range, ranking criterion). Use describe_data to confirm a field exists.
RATE_LIMITEDYou hit your tier's per-minute or per-day cap.Wait briefly, or check get_account_status for your exact limits.
INSUFFICIENT_CREDITSYour credit balance is exhausted.Top up or upgrade at mrmarket.ai. Not a fault, no retry needed.
SCHEMA_DRIFTA data field is temporarily unavailable (mid-update).Rephrase without the affected scope and try again.
INVALID_CURSOR / SNAPSHOT_EXPIREDA fetch_page cursor was malformed or its result snapshot aged out (snapshots live ~15 minutes).Re-run the original query_data question to get a fresh result and cursor.
BILLING_UNAVAILABLE / SERVICE_ERRORA transient server-side issue.Retry once; the retryable flag will be set.
Empty result on a screenFilters were too tight.Loosen one threshold (e.g. ROIC > 15% instead of 25%) or widen the time range.
A result looks wrongAn assumption or default may not match your intent.Read the assumptions array. If it still looks wrong, file it with report_issue (include the query_id).

For a guided tour of capabilities, limits, and worked examples at any time, call the free getting_started tool.

Support

  • Report a wrong result, bug, or feature request with the report_issue tool. Pass the query_id from the response so the exact run can be investigated. This is the fastest path because it links directly to the query trace.
  • Email: [email protected]

Links

来源：README.md，提交 df706d8

工具

0
工具元数据尚未被收录。

版本历史

1
  1. v1.1.0最新Sep 16, 2026