cancel_research_request
ChatGPTCancel one owned queued or running research request and release its unspent reservation. Free. A completed paid response remains available and is not refunded or deleted. This changes only the caller’s transient research request; it does not change archive data, Lists or other accounts.
describe_archive
ChatGPTUse when checking what readable evidence exists for a requested UTC period and scope. metrics=counts returns exact post and author counts; coverage adds asset, source, embedded and engagement diagnostics. This is not a mandatory first call, complete source history or a live collection watermark.
discover_evidence
ChatGPTDiscover all attached candidate identities and separately available unattached matching evidence for explicit stored-data filters. Discovery wording does not define a sector. Use identity-wide measurement for total project discussion; use exact-post selection for keyword-only facts. Source interpretation remains with you. Connected V2 research. A successful data operation uses the shared credit ledger; retry identity is explicit.
get_list
ChatGPTUse when a selected List’s exact assets, authors or saved references are needed for follow-up research. Returns bounded membership, pins and saved-query references with explicit truncation. Query List activity with query_list_activity; saved-query execution and edits are not implemented. Follow returned page arguments and revisions for complete requested directory metadata. Activity research does not require inspecting List membership first.
get_profile
ChatGPTIdentify the CryptoSocials account connected to this app. Free authenticated read of its stable account ID, display name and email. Use for account selection and connection help, not crypto research. It never selects or replaces another account.
get_research_balance
ChatGPTRead the account’s shared research-credit allowance, used/pending/remaining credits and period. Free; no research calculation. Use balance when the user asks for current credits, allowance/reset or a budget check; before a substantial research expansion when allowance matters; or when shared spending, a funding-period change or a refusal makes the balance uncertain. During ordinary research and follow-ups, reuse a recent balance and actual returned charges. Do not routinely check before or after every request, or poll while idle. A locally updated balance is only an estimate: other connections and pending reservations may change available credits. Respect explicit user spending limits and refresh when a decision needs current allowance. Balance totals cannot establish an individual uncertain charge; use recent/status/recover for uncertain retained-research delivery.
get_research_capabilities
ChatGPTRead the connected research guide, supported capabilities and evidence interpretation. Free; no research calculation or provider request.
get_research_guide
ChatGPTUse when asked what CryptoSocials can investigate, for research ideas, about its evidence or how research credits work. Free guidance, not a research request. Overview introduces combinable topics; choose a topic for more detail. A specific question can proceed directly without this call.
get_research_status
ChatGPTRead one owned research operation’s state, recorded credits and whether exact recovery is available. Free; no new calculation. Do not poll idle work.
get_usage
ChatGPTUse when checking research-credit balance, spending or a requested research budget. Free read of current account limits in credits.limits, separate transport_limits, allowance, used/pending/remaining credits, allowance period, operation costs and paged receipts. All keys, MCP, REST and Archive share the balance. Period end is not guaranteed renewal. Receipts are not conversations or complete question costs; the current call is recorded after preparation. Use balance when the user asks for current credits, allowance/reset or a budget check; before a substantial research expansion when allowance matters; or when shared spending, a funding-period change or a refusal makes the balance uncertain. During ordinary research and follow-ups, reuse a recent balance and actual returned charges. Do not routinely check before or after every request, or poll while idle. A locally updated balance is only an estimate: other connections and pending reservations may change available credits. Respect explicit user spending limits and refresh when a decision needs current allowance. Balance totals cannot establish an individual uncertain charge; use recent/status/recover for uncertain retained-research delivery.
list_lists
ChatGPTUse when the user wants to research the coins or Crypto Voices they follow. Returns the connected account’s saved List directory, ordered by UUID. If there is no List, ask for their named interests. Does not modify Lists or advance seen positions. Follow returned page arguments and revisions for complete requested directory metadata. Activity research does not require inspecting List membership first.
list_market_series
ChatGPTUse when choosing the stored price source for an exact asset and UTC period before comparing posts with prices. Returns series identities and available candle intervals. Sources remain separate; this does not select a winning price or fetch missing history. Follow returned page arguments and revisions for complete requested directory metadata. Activity research does not require inspecting List membership first.
list_research_requests
ChatGPTRead bounded recent request metadata for this account and connection permission. Match a request or operation ID to its recorded state and credits after an uncertain response. Free; no source text or new calculation.
measure_evidence
ChatGPTCalculate the complete accepted stored-data population across explicit periods, identities or exact selected posts. Use for numeric-only summaries, observed engagement, grouping, comparisons or numeric top posts (source_post grouping, metric order and top with all cutoff ties). When the question requires reading every selected original plus per-author or post-type counts, one complete read_evidence supplies those facts; do not measure the same population first. Delivery size never changes totals. result.identities already maps entity/id/label/symbol; reuse these UUIDs for follow-up filters instead of rediscovering the same projects. Missing observations remain unknown; arithmetic does not establish thematic relevance. Connected V2 research. A successful data operation uses the shared credit ledger; retry identity is explicit.
read_evidence
ChatGPTFor every selected original, chronology or positions plus per-author/post-type counts, start with one complete fresh reading; finish all selected pages, combine proven reused rows, and derive those counts from the selected posts. No separate measurement precheck. For an exact X post link use scope.select={kind:"post_urls",urls:[link]}, reading.blocks=[{id:"selected",select:{kind:"everything"}}], projection="reading"; no dates or author lookup required. Read exact versions, counters, relationships and context. The reading projection automatically includes available attached_prices per post; [] means none in those captured facts. No include_prices or include_attached_prices option exists. delivery.total counts selected posts before paging unless reading_reuse is present; then reading_reuse.selected_posts is the full selection and delivery.total counts the new/changed delivery. Context and overlapping-block duplicates never add to selected counts. A subset total is not a whole-scope total. Use next.arguments; reuse delivered wording. Outside-period or other-coin originals remain context. Connected V2 research. A successful data operation uses the shared credit ledger; retry identity is explicit.
read_price_context
ChatGPTRead a chosen retained price series and interval with actual observations, timing, provenance and gaps. For stored provider candles or derived summaries use view=candles, candle_basis=stored, time_axis=price_effective and interval seconds from inventory. Default point_samples candles use only individual points. Attached ingestion prices do not automatically establish publication prices. You choose and interpret the event; no upstream fetch, market-move threshold or inferred outcome. Connected V2 research. A successful data operation uses the shared credit ledger; retry identity is explicit.
recover_research_result
ChatGPTRecover one owned ready operation’s exact previously paid response after checking current permission and source access. Free; never fresh research, another page or a new charge. Continue further pages with the returned data tool arguments.
resolve_assets
ChatGPTUse when a user names a coin, ticker or known contract and research needs its exact identity. Query one name OR symbol (BTC, $BTC and #BTC supported), canonical key or exact stored namespaced contract key. Confirm the intended candidate; a symbol alone is not proof. If a combined name/ticker finds none, retry each separately. Unknown/merged states stay explicit. Follow returned page arguments and revisions for complete requested directory metadata. Activity research does not require inspecting List membership first.
search_authors
ChatGPTUse when the user wants to research an X author by handle or name. Find candidates by platform user ID, handle prefix or stored display name, then use get_author for profile, posts and activity. Name matches may be historical; confirm the intended identity. A successful exact missing @handle lookup can record the handle locally for coverage review; no provider call or promise of addition. Follow returned page arguments and revisions for complete requested directory metadata. Activity research does not require inspecting List membership first.
