research_briefing
ChatGPTUse this when the user requests the latest Market Update show, its desk briefing or chapters, e.g. 'What did today's show cover?'. Do not use for live market rankings (research_markets), an arbitrary historical show, Daily News or full archive search. Selection: No arguments: selects the latest actually published edition, which may not be dated today. Use its actual publication time. Presentation: Published briefing and chapter visuals. Refer to the relevant chapter and exact playback link; do not claim video is playing or embedded unless the host confirms it. Evidence: Use /edition and its named children for exact published text/charts. Show figures belong to that edition, not the current market.
research_capabilities
ChatGPTUse this when the user asks what can be investigated, or the assistant needs to resolve which workflow, metric, widget or control serves a question. Do not use for live data availability, mandatory preflight before familiar tools, general financial definitions already understood, or rendering another widget. Selection: Filter by area, short keywords or exact use_case_id. All query words must match; page using returned next_offset. Empty query lists that area. Broaden keywords if empty. Presentation: Structured help, not a market result. Explain relevant possibilities and why they help; distinguish tool actions from local controls and unsupported requests. Evidence: Cases are versioned descriptions of implemented workflows, not observed market evidence. Every case gives a question, inputs, tool sequence, visual role, explanation and boundary. No Research API read occurs.
research_cohort_status
ChatGPTUse this when a prior research_prepare_cohorts result is preparing and supplies a job_id. Do not use for starting an unrelated job, refreshing a ready comparison, or rapid/unbounded polling. Selection: Pass that exact returned job_id. Stop polling on ready, failed, unavailable or expired results; respect retry_after_seconds when provided. Presentation: Progress or the ready comparison, not another unrelated analysis. When ready, apply the user's groups and filters with research_compare_cohorts. Evidence: Reuse returned snapshot_id and inspect the returned catalogue/selection. Do not imply a preparation default answers the requested comparison.
research_compare_cohorts
ChatGPTUse this when measuring group performance, relative returns, contributions, metric relationships or qualified pair correlation, including refinements of an existing comparison. Do not use for inventing sector membership, re-preparing history unnecessarily, causal claims or portfolio advice. Selection: Use returned snapshot_id, exact catalogue group IDs and second-based bounds inside that history. Preserve unchanged selection fields, especially benchmark/overlap/filters. Request correlation with an explicit asset pair and supported window. Presentation: Use full category names, e.g. decentralised finance (DeFi) and artificial intelligence-related crypto projects (AI). State the period, measured metric and benchmark, then explain why the observed difference matters. Explain shared-member treatment briefly. Inspect member/contribution evidence before claiming a move was broadly shared or concentrated. Evidence: Use /lines, /members, /correlation and returned source/membership metadata. API analytics are authoritative; endpoint differences are not correlations.
research_discover
ChatGPTUse this when resolving an asset name/symbol or checking governed taxonomy membership, e.g. 'Which assets belong to DeFi?'. Do not use for performance rankings (research_markets) or measured group relationships (research_compare_cohorts). Selection: Search names/tickers with query. Use an exact returned group scope_id for group_id; reuse known IDs. Paging filters assets, not group labels. Presentation: Asset-directory results; use full category names and briefly explain unfamiliar terms. Open a ticker chart only if it serves the question. Evidence: Read /assets and /groups via research_read_evidence to obtain names, tickers and group IDs; counts alone do not identify members.
research_interpret_range
ChatGPTUse this when the user asks what a specific A/B chart selection means across its enabled metrics. Do not use for a whole-market briefing, predictions, endpoint-only narration or recovering the exact previous chart read. Selection: Use actual A/B candle opening times, timeframe, captured as_of, selection.alignment=candles, all selected layers and exact study IDs/code/calculation/timeframe. Calculated study timeframe must match A/B; both endpoint candles are included. Presentation: Deterministic prominent-move findings from Research, not MIA prose or an LLM call. Explain relevance accessibly without restating every metric. Evidence: This is a fresh history read, not the frozen A/B table. Returned range_end_exclusive and evidence_read_at govern it; /metric_checks provides detail.
research_market_context
ChatGPTUse this when the user asks about an asset's model regime, changing volatility or how unusual its volatility is within its own history. Do not use for price predictions, trading decisions, calibrated confidence, annualised volatility or the retired deterministic regime archive. Selection: Use the resolved Research ticker and exact timezone-qualified window, at most 31 days. Every observation is at hourly close regardless of the price chart interval. Presentation: A focused regime and volatility view. Keep 4-hour, 1-day and 7-day magnitudes separate; historical percentiles compare prior observations of this same market. Evidence: Read latest, current_status, processor_status, source pair, coverage and publication times before interpretation. Null ranks may mean building history even when a volatility magnitude exists. USD display may proxy the observed USDC pair; preserve both identities.
research_markets
ChatGPTUse this when the user wants the prepared market board or assets ranked for analytical attention, e.g. 'Which charts deserve a closer look?'. Do not use for published show content (research_briefing), a biggest-gainers ranking, or buy/sell recommendations. Selection: query filters ticker symbols; offset/limit page the existing published order, they do not sort it. Prefer a small requested shortlist. Presentation: Market-board rows with source clocks; explain Signal Book rank as attention, not expected return. Evidence: Use /items for exact rows and /metric_metadata for their clocks/meaning. A page is not the full market population.
research_news_context
ChatGPTUse this when the user asks what was reported about an asset around a day, or wants news alongside the chart. Do not use for video chapters, exhaustive web search, or treating mention volume as a count of source-linked articles. Selection: Use the resolved ticker and explicit anchor date. The API determines and returns the actual news/lookback window; this tool cannot request an arbitrary A/B news range. Presentation: Source-linked news and daily topic context. Explain what the coverage adds; do not present publication timing as a causal explanation for price. Evidence: Retrieve relevant returned sections before citing titles/counts. Respect date-only timestamps and distinguish corpus mentions from linked stories.
research_open_ticker
ChatGPTUse this when opening an asset's native price chart and historical replay for a specified interval, e.g. 'Show BTC over the past week'. Do not use for adding a metric to an already-selected chart (research_read_series), automatic interpretation, or a live execution quote. Selection: Use the resolved ticker and timezone-qualified window; end is exclusive. Reuse the requested candle interval. Resolve ambiguous dates first. Presentation: Data result first; research_show_price displays a focused price chart with local replay and A/B selection. Requested metrics use separate focused widgets. Do not add every metric by default. Evidence: Exact rows: /candles/rows, including source-minute coverage. Layer catalogue: /catalogue/datasets. Selected-interval calculated study IDs/clocks: /indicators/items. Raw rows are widget-only until retrieved.
research_prepare_cohorts
ChatGPTUse this when starting a group relationship comparison without a usable prepared snapshot, e.g. comparing decentralised-finance and artificial-intelligence crypto tokens over seven days. Do not use for taxonomy lookup alone or re-preparing a suitable snapshot for every filter change or replay frame. Selection: days selects 1, 7 or 30 days of retained history. Resolve requested group IDs from the returned catalogue, not guessed labels. Presentation: Preparation/initial comparison view. The initial default groups are not necessarily the user's groups; call research_compare_cohorts with their selection once ready. Evidence: While preparing, use job_id with research_cohort_status and respect retry guidance. Once ready retain snapshot_id; inspect /initial/catalogue, /initial/selection and /initial/history_range when present.
research_read_evidence
ChatGPTUse this when the assistant needs exact values, catalogue IDs or source detail from a result already returned by another Research tool. Do not use for refreshing data, changing the selection, raw API access or opening another widget. Selection: Use the relevant result's evidence.id and a JSON-pointer path. Inspect the root for child names if unknown; request bounded arrays with offset/limit and follow next_offset. Presentation: Structured evidence only; no dedicated widget. Use the result to answer or configure a follow-up, without dumping raw JSON at the user. Evidence: Objects return fields/children, arrays return items/total/next_offset, scalars return value. Same captured read, identity-isolated and expiring; an expired reference is not silently refreshed.
research_read_series
ChatGPTUse this when the user asks for one metric, study, more source rows, or a follow-up such as 'Add funding to this chart'. Do not use for computed range interpretation (research_interpret_range), invented indicators, or replay animation frames. Selection: Retain ticker, window and timeframe. Indicators default to calculation=candles at 1h/4h/1d with exact variant ID/code. Numeric snapshot/flow families default to alignment=candles at that interval. Native is an explicit raw-source selection, never a fallback. Preserve the exact query when following next_cursor. Presentation: Use facts.display_tool for its focused widget. Alignment and flow sums belong to Research; retain units and observed_at/available_at. OI lines use the API's nested matching-cycle venues, never local sums. Pattern events have no widget. Evidence: Read /rows and units/timing/coverage/selected_study before explaining values. Null readings are not zero. Inspect candle_end/readiness/usable/comparison_eligible; partial flow subtotals are not complete totals. next_cursor means this is not all upstream history.
research_show_briefing
ChatGPTDisplay a focused published market update widget after research_briefing has returned the requested evidence. A published edition and its chapters. Published figures are not current market quotes. Use only when a visual helps answer the question. Not a data fetch, calculation, navigation hub or replacement for the assistant's interpretation. Reuses captured source numbers and selection; expired evidence requires an explicit new read.
research_show_cohorts
ChatGPTDisplay a focused group comparison widget after research_compare_cohorts has returned the requested evidence. One pinned comparison: performance, benchmark-relative performance, member contributions, relationships or qualified correlation. Preserve named groups and selection. Use only when a visual helps answer the question. Not a data fetch, calculation, navigation hub or replacement for the assistant's interpretation. Reuses captured source numbers and selection; expired evidence requires an explicit new read.
research_show_directory
ChatGPTDisplay a focused asset directory widget after research_discover has returned the requested evidence. A compact list of resolved assets and their named category memberships. Use only when a visual helps answer the question. Not a data fetch, calculation, navigation hub or replacement for the assistant's interpretation. Reuses captured source numbers and selection; expired evidence requires an explicit new read.
research_show_funding
ChatGPTDisplay a focused funding chart widget after research_read_series has returned the requested evidence. Funding history in its supplied units. Positive rates mean longs pay shorts; negative rates reverse that payment. Use only when a visual helps answer the question. Not a data fetch, calculation, navigation hub or replacement for the assistant's interpretation. Reuses captured source numbers and selection; expired evidence requires an explicit new read.
research_show_liquidations
ChatGPTDisplay a focused liquidation chart widget after research_read_series has returned the requested evidence. Long and short liquidation flows over the selected period, with separate direction labels. Use only when a visual helps answer the question. Not a data fetch, calculation, navigation hub or replacement for the assistant's interpretation. Reuses captured source numbers and selection; expired evidence requires an explicit new read.
research_show_liquidity
ChatGPTDisplay a focused order-book liquidity widget after research_read_series has returned the requested evidence. Displayed bid/ask depth or quoted spread. Depth and spread use separate scales, not a misleading combined axis. Use only when a visual helps answer the question. Not a data fetch, calculation, navigation hub or replacement for the assistant's interpretation. Reuses captured source numbers and selection; expired evidence requires an explicit new read.
research_show_market_context
ChatGPTDisplay a focused market regime and volatility widget after research_market_context has returned the requested evidence. Hourly model descriptions and separate 4-hour, 1-day and 7-day observed volatility. Keep readiness, gaps, source pair and historical percentiles explicit; probabilities are not calibrated confidence. Use only when a visual helps answer the question. Not a data fetch, calculation, navigation hub or replacement for the assistant's interpretation. Reuses captured source numbers and selection; expired evidence requires an explicit new read.
research_show_markets
ChatGPTDisplay a focused market shortlist widget after research_markets has returned the requested evidence. Ranked assets and their current reported prices; Signal Book rank measures analytical attention, not expected returns. Use only when a visual helps answer the question. Not a data fetch, calculation, navigation hub or replacement for the assistant's interpretation. Reuses captured source numbers and selection; expired evidence requires an explicit new read.
research_show_news
ChatGPTDisplay a focused news timeline widget after research_news_context has returned the requested evidence. Source-linked articles in the returned dated window. Topic mention counts are not article counts. Use only when a visual helps answer the question. Not a data fetch, calculation, navigation hub or replacement for the assistant's interpretation. Reuses captured source numbers and selection; expired evidence requires an explicit new read.
research_show_open_interest
ChatGPTDisplay a focused open interest chart widget after research_read_series has returned the requested evidence. USD open interest by reported venue. Keep separate venue lines; not a locally invented aggregate. Use only when a visual helps answer the question. Not a data fetch, calculation, navigation hub or replacement for the assistant's interpretation. Reuses captured source numbers and selection; expired evidence requires an explicit new read.
research_show_positioning
ChatGPTDisplay a focused positioning chart widget after research_read_series has returned the requested evidence. Long/short account ratios or tracked-position observations, without treating account ratios as capital flows. Use only when a visual helps answer the question. Not a data fetch, calculation, navigation hub or replacement for the assistant's interpretation. Reuses captured source numbers and selection; expired evidence requires an explicit new read.
research_show_price
ChatGPTDisplay a focused price chart widget after research_open_ticker has returned the requested evidence. Candlestick history with local 3× replay and A/B selection. Ask the assistant to investigate the selected move. Use only when a visual helps answer the question. Not a data fetch, calculation, navigation hub or replacement for the assistant's interpretation. Reuses captured source numbers and selection; expired evidence requires an explicit new read.
research_show_study
ChatGPTDisplay a focused technical study widget after research_read_series has returned the requested evidence. One requested candle-calculated indicator with exact identity, interval and period, or an explicitly native indicator. Retired regimes and pattern events are not exposed. Use only when a visual helps answer the question. Not a data fetch, calculation, navigation hub or replacement for the assistant's interpretation. Reuses captured source numbers and selection; expired evidence requires an explicit new read.
research_show_videos
ChatGPTDisplay a focused video timeline widget after research_ticker_videos has returned the requested evidence. Ticker-matched published chapters with exact watch links and dates. Use only when a visual helps answer the question. Not a data fetch, calculation, navigation hub or replacement for the assistant's interpretation. Reuses captured source numbers and selection; expired evidence requires an explicit new read.
research_ticker_videos
ChatGPTUse this when the user wants a ticker's published video chapters on a selected timeline. Do not use for unrelated market-wide videos, general web video search or proof that coverage caused a price move. Selection: Use the resolved ticker and publication-time window; end is exclusive. Preserve the active ticker/date context on follow-ups. Presentation: Ticker-primary coverage timeline with exact chapter links. Keep publication time distinct from the market period discussed. Evidence: Inspect returned sections for titles, matching basis and timestamps before describing an item; do not infer relevance merely from its publication date.
research_briefing
Clauderesearch_capabilities
Clauderesearch_cohort_status
Clauderesearch_compare_cohorts
Clauderesearch_discover
Clauderesearch_interpret_range
Clauderesearch_market_context
Clauderesearch_markets
Clauderesearch_news_context
Clauderesearch_open_ticker
Clauderesearch_prepare_cohorts
Clauderesearch_read_evidence
Clauderesearch_read_series
Clauderesearch_show_briefing
Clauderesearch_show_cohorts
Clauderesearch_show_directory
Clauderesearch_show_funding
Clauderesearch_show_liquidations
Clauderesearch_show_liquidity
Clauderesearch_show_market_context
Clauderesearch_show_markets
Clauderesearch_show_news
Clauderesearch_show_open_interest
Clauderesearch_show_positioning
Clauderesearch_show_price
Clauderesearch_show_study
Clauderesearch_show_videos
Clauderesearch_ticker_videos
Claude
