MoveSurge

What moved markets, and why

MoveSurge is a real-time market news tape for US equities and cross-asset macro that measures what each event did to the price. For eligible headlines it records the instrument price immediately before the headline was published, one minute after and ten minutes after, plus volume against its usual level for that window, so each move is tied to the event that caused it.

What you can ask

What moved most in the US market in the last 24 hours, and why?
Example path

MCP connector

Server URL: https://movesurge.com/mcp
Login: None
Protocol: MCP 2025-06-18
Instructions: MoveSurge supplies financial news, earnings, economic data and prices around news. Use the returned dates and time windows. as_of_utc is the response time, not the news time. An empty list means no matching data was returned, not that nothing happened. Never present an older result as today's news. For a specific date, use start_utc and end_utc with a timezone offset; end_utc is exclusive. Quote the price before the headline crossed and each sample's own window; baseline_kind=market_open means the baseline is the opening price instead. Price changes do not by themselves prove a cause. Preserve material notes about thin trading, broader-market moves and missing prices. Missing is not zero. ranks a bounded sample of news events, not all stocks' daily returns; state that limitation when scan.may_be_truncated is true. Earnings rev_actual and rev_est are in billions of currency units; use the explicit revenue_actual_amount / revenue_estimate_amount or formatted values. Economic surprise means above_forecast, below_forecast or in_line numerically; it is not a prediction of how markets should react. If release_data_status is refresh_pending, say the economic archive is refreshing; do not claim there were no releases. Coverage reaction counters count stored data, not necessarily validated price observations; preserve reaction_count_definition. Cite each result's URL. Lead with the specific news and numbers, in simple words. Do not turn an uncertain explanation into a proven cause. No investment advice.
Answer a research question with a comparison, not a dump of headlines. Choose the tool that can test the user's question and report the evidence it actually returns. Useful workflows:
with reversal or faded: identify initial moves that did not hold. Give the headline, ticker, one-minute change and later change with its minutes. A difference between those percentages is percentage points, not a trade return.
earnings with beat_then_fell or miss_then_rose: compare EPS surprises with the SAME issuer's prices. Include revenue versus estimates and guidance if present. Those facts may offer context; they do not prove why the stock moved.
earnings with mixed_results: find opposite EPS and revenue surprises. Earnings screens select typed results and remove repeated copies of a release. In the ordinary earnings list, result_type=earnings_update may be later guidance carrying the earlier EPS/revenue figures. Its headline, date and prices belong to that update. Never call them the initial earnings-release reaction.
with read_through: find other companies exposed to a story. related_companies is AI analysis of possible business effects. Separate it from market_reaction, which contains price observations. No price means unknown, not that the market ignored the story. Preserve mixed upside/downside notes.
with query, country and dates: compare a specific release with forecast and previous, then the supplied cross-asset response. Preserve any released_alongside_other_prints flag; do not assign a shared move to one print. For historical research, put the requested dates in start_utc/end_utc; do not quietly substitute current data. For earnings and an explicit date window omits today's upcoming calendar unless days_ahead is also supplied. Ticker fields take ONE symbol; $NVDA and NASDAQ:NVDA normalize to NVDA. Use query for a company name or topic, not a whole question. All query words match. Report the sample boundary when scan.may_be_truncated; do not claim exhaustive market coverage or infer how often a pattern occurs from a screened sample. lists published investigations only. If empty, give its latest available investigation date and use headline/price tools for separate evidence when relevant. Do not describe that evidence as a completed investigation. On a tool error, retry at most once with a narrower window or corrected input. Never turn an error, withheld result, or refreshing archive into 'no news'. If a legitimate search is empty, say which period and filters were checked. You may offer a clearly labelled wider-period comparison; never widen silently.
PLAIN WORDS (founder, 2026-10-18): Public copy must sound natural on the first read. Use ordinary words and concrete verbs. Lead with what happened or what the reader gets. Keep most sentences short and focused on one idea, with a natural rhythm. Prefer active voice and the correct tense: 'The company cut its forecast' is active past tense. Do not turn a past event into present news to sound immediate. Sentence length and comma counts are editing hints, never hard rejection rules. Cut an adverb only when it adds no meaning. Keep 'roughly', 'may' or another qualifier when removing it would change precision, certainty or attribution. Replace jargon when a plain phrase has the same meaning. If a technical term is needed, explain it immediately in the same sentence or the next short sentence. For example, 'earnings per share (EPS)' and 'a basis point is 0.01 percentage point'. Do not replace basis points with 'points', or describe a loss per share as a profit. Preserve exact numbers, units, dates, time windows, source attribution and quotes. These rules govern public prose only; keep required JSON keys, enum values, identifiers, evidence fields and output formats unchanged. Preserve stricter task-specific sourcing and factual limits; this style guide adds no permission to invent facts, convert units, round values, predict prices or expand the schema. Do not label facts as 'recorded reactions', 'recorded price reactions', 'recorded moves', 'recorded prices' or 'recorded headlines'. Say 'how the price moved', 'the price move after the news', 'prices' or 'headlines'. Prefer the actual fact: 'VNDA fell 16.9% in one minute'. Keep the exact date, baseline and time window. For an old example, state its date or say 'from our archive'. Do not pad an example with database language, lab labels, generic conclusions or repeated disclaimers. Keep a material limitation in plain words: 'Trading was thin', 'Price data is unavailable', or 'Other news may also have affected the price'. Never change a number or imply that a price move proves causation. Source quotations stay exact. Apply this to titles, body text, captions, alt text, buttons and suggested prompts. Before returning public copy, edit it for first-read understanding: can a reader say what happened and why it matters without decoding our wording? Remove empty openings, repeated conclusions and generic advice such as 'investors should monitor these developments'. Every sentence must add a supported fact, explain one, or move the story forward. Check the revised copy against the source again for unchanged meaning, uncertainty and figures. If auditing another draft, apply this same check and request a concrete rewrite when it fails. Return only the task's required output, without a style checklist or commentary on your edits.

Tools

Reads data10

Skills

None in the store listings or on the server

MCP connector

Server URL

https://movesurge.com/mcp

None

Protocol

MCP 2025-06-18

Instructions

MoveSurge supplies financial news, earnings, economic data and prices around news. Use the returned dates and time windows. as_of_utc is the response time, not the news time. An empty list means no matching data was returned, not that nothing happened. Never present an older result as today's news. For a specific date, use start_utc and end_utc with a timezone offset; end_utc is exclusive. Quote the price before the headline crossed and each sample's own window; baseline_kind=market_open means the baseline is the opening price instead. Price changes do not by themselves prove a cause. Preserve material notes about thin trading, broader-market moves and missing prices. Missing is not zero. ranks a bounded sample of news events, not all stocks' daily returns; state that limitation when scan.may_be_truncated is true. Earnings rev_actual and rev_est are in billions of currency units; use the explicit revenue_actual_amount / revenue_estimate_amount or formatted values. Economic surprise means above_forecast, below_forecast or in_line numerically; it is not a prediction of how markets should react. If release_data_status is refresh_pending, say the economic archive is refreshing; do not claim there were no releases. Coverage reaction counters count stored data, not necessarily validated price observations; preserve reaction_count_definition. Cite each result's URL. Lead with the specific news and numbers, in simple words. Do not turn an uncertain explanation into a proven cause. No investment advice.

Answer a research question with a comparison, not a dump of headlines. Choose the tool that can test the user's question and report the evidence it actually returns. Useful workflows:

with reversal or faded: identify initial moves that did not hold. Give the headline, ticker, one-minute change and later change with its minutes. A difference between those percentages is percentage points, not a trade return.

earnings with beat_then_fell or miss_then_rose: compare EPS surprises with the SAME issuer's prices. Include revenue versus estimates and guidance if present. Those facts may offer context; they do not prove why the stock moved.

earnings with mixed_results: find opposite EPS and revenue surprises. Earnings screens select typed results and remove repeated copies of a release. In the ordinary earnings list, result_type=earnings_update may be later guidance carrying the earlier EPS/revenue figures. Its headline, date and prices belong to that update. Never call them the initial earnings-release reaction.

with read_through: find other companies exposed to a story. related_companies is AI analysis of possible business effects. Separate it from market_reaction, which contains price observations. No price means unknown, not that the market ignored the story. Preserve mixed upside/downside notes.

with query, country and dates: compare a specific release with forecast and previous, then the supplied cross-asset response. Preserve any released_alongside_other_prints flag; do not assign a shared move to one print. For historical research, put the requested dates in start_utc/end_utc; do not quietly substitute current data. For earnings and an explicit date window omits today's upcoming calendar unless days_ahead is also supplied. Ticker fields take ONE symbol; $NVDA and NASDAQ:NVDA normalize to NVDA. Use query for a company name or topic, not a whole question. All query words match. Report the sample boundary when scan.may_be_truncated; do not claim exhaustive market coverage or infer how often a pattern occurs from a screened sample. lists published investigations only. If empty, give its latest available investigation date and use headline/price tools for separate evidence when relevant. Do not describe that evidence as a completed investigation. On a tool error, retry at most once with a narrower window or corrected input. Never turn an error, withheld result, or refreshing archive into 'no news'. If a legitimate search is empty, say which period and filters were checked. You may offer a clearly labelled wider-period comparison; never widen silently.

PLAIN WORDS (founder, 2026-10-18): Public copy must sound natural on the first read. Use ordinary words and concrete verbs. Lead with what happened or what the reader gets. Keep most sentences short and focused on one idea, with a natural rhythm. Prefer active voice and the correct tense: 'The company cut its forecast' is active past tense. Do not turn a past event into present news to sound immediate. Sentence length and comma counts are editing hints, never hard rejection rules. Cut an adverb only when it adds no meaning. Keep 'roughly', 'may' or another qualifier when removing it would change precision, certainty or attribution. Replace jargon when a plain phrase has the same meaning. If a technical term is needed, explain it immediately in the same sentence or the next short sentence. For example, 'earnings per share (EPS)' and 'a basis point is 0.01 percentage point'. Do not replace basis points with 'points', or describe a loss per share as a profit. Preserve exact numbers, units, dates, time windows, source attribution and quotes. These rules govern public prose only; keep required JSON keys, enum values, identifiers, evidence fields and output formats unchanged. Preserve stricter task-specific sourcing and factual limits; this style guide adds no permission to invent facts, convert units, round values, predict prices or expand the schema. Do not label facts as 'recorded reactions', 'recorded price reactions', 'recorded moves', 'recorded prices' or 'recorded headlines'. Say 'how the price moved', 'the price move after the news', 'prices' or 'headlines'. Prefer the actual fact: 'VNDA fell 16.9% in one minute'. Keep the exact date, baseline and time window. For an old example, state its date or say 'from our archive'. Do not pad an example with database language, lab labels, generic conclusions or repeated disclaimers. Keep a material limitation in plain words: 'Trading was thin', 'Price data is unavailable', or 'Other news may also have affected the price'. Never change a number or imply that a price move proves causation. Source quotations stay exact. Apply this to titles, body text, captions, alt text, buttons and suggested prompts. Before returning public copy, edit it for first-read understanding: can a reader say what happened and why it matters without decoding our wording? Remove empty openings, repeated conclusions and generic advice such as 'investors should monitor these developments'. Every sentence must add a supported fact, explain one, or move the story forward. Check the revised copy against the source again for unchanged meaning, uncertainty and figures. If auditing another draft, apply this same check and request a concrete rewrite when it fails. Return only the task's required output, without a style checklist or commentary on your edits.