{"version":"2026-10-13.2","title":"Explore with your agent","introduction":"Ask about live Crypto X discussion, an author's history, a coin's changing story, or the posts around a price move. Combine people, assets, topics and dates, then follow up on what you find. Explore the ideas below or start with your own question.","url":"https://cryptosocials.net/research-guide","approach":"CryptoSocials supplies posts, identities, dates, source links and stored price observations. The research engine calculates facts across your requested period and filters; your agent reads the relevant evidence, makes comparisons and explains its findings. Ask for a quick catch-up or a deeper investigation, then follow the original sources and make any gaps clear.","topics":[{"id":"live","title":"Follow the discussion","summary":"See which coins people are discussing now and how attention is changing.","questions":["Which coins are attracting more discussion today, and what has changed?","Is $BTC attracting more voices this week, or repeated posts from the same few accounts?","What are people saying about the five most discussed coins today?","Which previously quiet coins have new discussion this week?","Who stands out in $SUI's community this week? Compare their posts, views and reposts."],"explore":["Compare equal time windows by post count and distinct authors.","Separate ordinary commentary from known repost activity, then read representative posts.","Break a period into days or weeks, or narrow it to selected authors and assets.","Follow a ranking into a coin's posts, author claims and price timeline."],"evidence":"Whole-window counts, dated posts, author identities, source links and available engagement observations.","tools":["discover_activity","aggregate_posts","compare_activity","search_posts","get_posts","list_live_views","get_live_summary"],"workflow":"For topic discovery and factual comparisons, use discover_activity when advertised. Supply the discovery search separately from current measurement filters and an explicit baseline when growth is requested. It measures every discovered candidate before pagination; inspect remaining candidate pages and relevance sources for a thematic answer. For requested leading N in the same numeric order, ranking_bound permits stopping only after verifying every candidate that could outrank the chosen N and the lowest chosen value strictly exceeds the highest unread value. Resolve or disclose ties; narrative and sector completeness still require context checks. Reuse the returned counts and separately labelled period citations. With older definitions, paginate aggregate_posts discovery results, reuse whole_identity_activity without discovery text, then compare every relevant candidate with compare_activity for growth. Use aggregate_posts with metrics=counts and the requested order for full-window rankings. A leading N result does not require reading lower-ranked groups. compare_activity measures every selected coin or author across both periods, including known zeros; union totals are separate. Read selected full posts before explaining what changed. Archive engagement supports the same combined filters and account time limits, up to 366 days; Live engagement remains one coin and seven days. Rank summed counters separately from highest single-post counters. get_asset/get_author also return a first-page engagement_summary across the whole requested period, including numeric availability. Use it before scanning posts, especially with an older installed tool definition. Preserve dates, filters, attribution and source links.","context":"These are collected-post counts, not all of X. Popularity is not sentiment or prediction quality. Live summaries show a limited leading set; use aggregation for larger rankings. Views and other engagement totals are latest stored lifetime counters on ordinary posts (original and historical unknown labels), replies and quotes published in the chosen period. Numeric ingestion captures count as observed; missing fields stay unknown. Known repost targets are excluded from the reposting author’s engagement totals. Stored historical type labels remain unchanged. Views are not unique people or reach caused by that coin. A ranking does not establish support or influence."},{"id":"coins","title":"Understand a coin's story","summary":"Follow a coin across posts, people and time, with its exact identity attached.","questions":["What changed in the $SUI conversation this week?","What arguments are people making for and against $ETH this week?","Which $KAS voices have posts with the most views and reposts this week?","Compare the main developments and disagreements around $SOL and $ETH this week."],"explore":["Start from a ticker, name or known contract and confirm the intended coin.","Read a balanced selection across the period rather than only the latest posts.","Trace a claim through replies, quotes and later commentary where retained.","Compare discussion with the coin's stored price timeline."],"evidence":"Asset identities, matched posts, full-window activity, referenced posts and available price sources.","tools":["discover_activity","resolve_assets","get_asset","get_posts","get_post","list_market_series","get_market_history"],"workflow":"For topic discovery and factual comparisons, use discover_activity when advertised. Supply the discovery search separately from current measurement filters and an explicit baseline when growth is requested. It measures every discovered candidate before pagination; inspect remaining candidate pages and relevance sources for a thematic answer. For requested leading N in the same numeric order, ranking_bound permits stopping only after verifying every candidate that could outrank the chosen N and the lowest chosen value strictly exceeds the highest unread value. Resolve or disclose ties; narrative and sector completeness still require context checks. Reuse the returned counts and separately labelled period citations. With older definitions, paginate aggregate_posts discovery results, reuse whole_identity_activity without discovery text, then compare every relevant candidate with compare_activity for growth. Resolve the identity, then get_asset for posts, counts and whole-period engagement summaries together. Read selected full bodies before attributing claims and cite their source links. Use separately dated market history when the question needs prices. Reuse the first-page summary and bodies on follow-ups.","context":"A shared ticker can name several coins. Candidate matches stay explicit; a matching name alone is not identity proof."},{"id":"authors","title":"Research the people behind the posts","summary":"Look beyond one successful call: inspect an author's claims, changes of view and track record in context.","questions":["Has @CryptoKaleo changed his view on $ETH this week?","Compare @CryptoKaleo and @VuoriTrading on $BTC this month. Where do their arguments agree or differ?","Find a specific price claim from @CryptoKaleo this week. How much of the move had already happened when he posted?","Does @CryptoKaleo focus on a few clear theses or discuss many different coins?","What evidence supports and challenges @CryptoKaleo's recent $ETH argument?","Which of @CryptoKaleo's posts this week attracted the most views and sharing?"],"explore":["Inspect the actual wording, date, entry conditions, targets and invalidations.","Compare an author's earlier and later views on the same asset.","Check other calls and misses before judging a selected success.","Look at posting breadth, frequency and dated audience/engagement observations.","Choose a relevant cohort, explain your comparison criteria and follow promising examples into their sources."],"evidence":"Stored author profiles and handles, post bodies, whole-window counts, engagement checkpoints and price observations.","tools":["discover_activity","search_authors","get_author","aggregate_posts","get_posts","get_engagement_history","list_market_series","get_market_history"],"workflow":"For topic discovery and factual comparisons, use discover_activity when advertised. Supply the discovery search separately from current measurement filters and an explicit baseline when growth is requested. It measures every discovered candidate before pagination; inspect remaining candidate pages and relevance sources for a thematic answer. For requested leading N in the same numeric order, ranking_bound permits stopping only after verifying every candidate that could outrank the chosen N and the lowest chosen value strictly exceeds the highest unread value. Resolve or disclose ties; narrative and sector completeness still require context checks. Reuse the returned counts and separately labelled period citations. With older definitions, paginate aggregate_posts discovery results, reuse whole_identity_activity without discovery text, then compare every relevant candidate with compare_activity for growth. Find the exact author with search_authors, then get_author for profile, posts, whole-period counts and engagement totals. Keep the exact returned handle attached to its measurements. Use aggregate_posts grouped by asset to examine their coverage. Verify selected claims and compare each with relevant subsequent prices; examine counterexamples as part of a performance question.","context":"The agent assesses the evidence; CryptoSocials does not assign a success score. A sampled comparison is not an exhaustive ranking of every author, and an old profile snapshot is not a current follower count."},{"id":"prices","title":"Connect posts with price history","summary":"Investigate what was said before, during and after a move.","questions":["What were people saying before $SOL’s largest price change this week?","Compare $BTC commentary this week with its price timeline. Who was early, who changed their view, and what did not hold up?","What warnings or conditions appeared in $SOL commentary this week, and what happened afterward?","Find a price target mentioned for $SUI this week. How far away was it when posted, and what happened on the way?","Which explanations for $BTC’s price changes this week appeared beforehand, and which appeared afterward?"],"explore":["Inspect the timeline before choosing the relevant discussion window.","Compare specific claims with later observations, including intervening adverse moves.","Separate conditions and proposed entries from claims that a trade actually happened.","Investigate conflicting explanations and counterexamples."],"evidence":"Original post times, selected stored price sources, raw points or available candles, source links and observation dates.","tools":["resolve_assets","list_market_series","get_market_history","get_asset","get_posts","get_price_context"],"workflow":"Resolve the asset, then read relevant posts with projection=market to obtain full wording and attached price evidence together. Start a specific claim with publication_market_context, checking its method and age; market_context is the separate ingestion capture. Reuse a returned series_id for get_market_history, or list sources when none is known. For a post outcome, start the history at that post’s publication time and end at the chosen cutoff. A whole-week high or low may precede the post. Continue raw points only when the sequence needs it; a window summary alone cannot establish target order or conditional entry. Preserve conditions, invalidations and gaps.","context":"Post time, price-effective time and capture time are different. Gaps stay visible. Nearby observations are not exact trade prices, and sequence alone does not prove a post caused a move."},{"id":"chains","title":"Explore connected assets","summary":"Narrow research by stored chain associations and compare the assets within them.","questions":["Which Base-associated tokens are getting more discussion this week?","What are people actually saying about the leading Solana-associated assets this week?","Which multichain assets appear in both Base and Ethereum research this week?","Compare discussion of Solana with tokens associated with its chain this week."],"explore":["Filter by a known chain, native/token relationship or multichain status.","Rank associated assets, then investigate selected coins and their actual posts.","Combine chain filters with authors, dates or text topics.","Inspect the identity details when an association is surprising."],"evidence":"Confirmed stored asset-chain associations, asset identities, readable posts and activity counts.","tools":["aggregate_posts","search_posts","get_asset","resolve_assets"],"workflow":"Use chain_ids and, when relevant, asset_type or multichain on supported search/aggregate tools. Keep each asset's actual association visible when explaining results.","context":"Association with a chain does not mean native to it. The same asset can occur in several chain scopes; a matching post is not necessarily about that chain. Category filters such as meme coins are not provided."},{"id":"lists","title":"Catch up on what you follow","summary":"Use your saved Lists to give research a personal starting point.","questions":["What did I miss from my List today?","Which of the voices I follow changed their view this week?","Pick three original posts worth reading from my List today and explain why.","Compare this week’s discussion with last week’s for the coins I follow."],"explore":["Choose a saved List and a period, or name the interests directly.","Read new discussion matching the List's assets or authors without duplicate posts.","Inspect related prices, original claims or conflicting views when useful.","Follow up on one finding without repeating the entire catch-up."],"evidence":"Your connected account's List names and members, matching posts, saved references and source links.","tools":["list_lists","get_list","query_list_activity","get_posts","compare_activity"],"workflow":"List active Lists, select the relevant one and query_list_activity for its deduplicated author-or-asset membership. Its first page returns exact requested-period and adjacent equal-period comparisons for each current member, with names and dates. Use those measurements instead of separate member rereads; a post page is not the period total. For a three-request catch-up use list_lists, get_list and query_list_activity with projection=reading, then answer from the returned bodies and comparisons. Reuse known List IDs and evidence on follow-ups. Archived Lists stay outside ordinary catch-up. If no List exists, research the user’s named interests instead.","context":"Only your connected account's Lists are available. These read tools do not edit Lists, mark posts seen or create alerts. Website controls manage those actions."},{"id":"history","title":"Trace a topic through time","summary":"Explore how a narrative developed and compare what different people said.","questions":["How has the $ETH scaling discussion changed over the past month?","Who challenged bullish $BTC arguments this week, and what reasons did they give?","Find a quoted $SUI claim this week. Was the quote agreeing, criticizing or simply sharing it?","Compare $SOL commentary this week with last week and show the original sources."],"explore":["Combine text, exact assets, authors and dates.","Follow retained reply, quote and repost relationships.","Compare periods and examine changes in both discussion volume and content.","Read engagement history using its observation dates, not just a present-looking total."],"evidence":"Searchable retained text, post relationships, provenance, publication dates and engagement checkpoints.","tools":["discover_activity","search_posts","get_post","get_posts","compare_activity","aggregate_posts","get_engagement_history"],"workflow":"For topic discovery and factual comparisons, use discover_activity when advertised. Supply the discovery search separately from current measurement filters and an explicit baseline when growth is requested. It measures every discovered candidate before pagination; inspect remaining candidate pages and relevance sources for a thematic answer. For requested leading N in the same numeric order, ranking_bound permits stopping only after verifying every candidate that could outrank the chosen N and the lowest chosen value strictly exceeds the highest unread value. Resolve or disclose ties; narrative and sector completeness still require context checks. Reuse the returned counts and separately labelled period citations. With older definitions, paginate aggregate_posts discovery results, reuse whole_identity_activity without discovery text, then compare every relevant candidate with compare_activity for growth. Start with a bounded, relevant period and exact identities where available. Read selected bodies and referenced context before interpreting stance. Narrow a costly broad-word search by dates or identities when needed, explaining incomplete coverage instead of calling a timeout an empty result. Text search supports spaced OR, quoted phrases and minus exclusions. Words do not stem: agent and agents differ. Repeat shared terms instead of grouping with parentheses, for example AI agent OR AI agents. A selected discovery shortlist is not every project; complete its requested comparisons and explain the remaining search scope.","context":"A repost is not an independent opinion, and a quote does not establish agreement. Retained text and relationships may be incomplete; absence from the archive is not proof nobody said it. Counters on known reposts cannot be assigned to the sharing author; captured original-target engagement remains separate. Cite the actual speaker and original source when available."},{"id":"coverage","title":"Understand coverage and credits","summary":"Check what evidence is available and choose a useful research scope.","questions":["How much research can I do with my remaining credits?","What can you investigate about $SUI with the evidence available?","Start with a small $BTC research pass and explain what a deeper one would add.","Can you answer a follow-up from the posts we already retrieved?"],"explore":["Check the actual shared balance without spending research credits.","Use existing results and source details before making duplicate requests.","Agree a bounded investigation when a broad question could require many pages.","Distinguish an empty result, missing history and a failed request."],"evidence":"Account allowance and operation costs, requested-window counts, price-source availability and explicit gaps.","tools":["get_usage","describe_archive","list_market_series","get_profile"],"workflow":"Use the free get_profile for connected-account questions. Use get_usage for balance or budget questions. Plan the data requests before starting and treat the user’s maximum as a hard stop. Identity lookups, successful empty results and extra pages each count. Reuse evidence already returned; stop with a useful partial answer and ask before exceeding the budget. describe_archive supplies counts/coverage for a requested scope, but is not a mandatory first call. Use whole-window summaries for numerical comparisons before reading pages. Agree expansions beyond the requested scope or an explicit spending limit. Pages needed to finish the accepted scope are part of the original investigation. Cite original source links beside findings and do not repeat identical successful reads.","context":"One credit currently pays for each successful data request, including empty results. Each extra page is another request. Connection/capability help and balance reads are free. A settled failed calculation releases its reservation. If delivery is interrupted, check the existing operation and recover a completed response freely before starting research again."}],"credits":{"explanation":"One research credit pays for one successful request for data. A question can combine several requests, and larger results can need more pages. MCP, REST and older website Archive research share your account balance. Website browsing within the current seven days is free; agent and API data requests retain their normal research-credit cost.","examples":[{"title":"A daily discussion ranking","credits":1,"includes":"One request for ranked coin counts across a selected day."},{"title":"An author's first research pass","credits":2,"includes":"Find an author, then read their first page of posts and the selected period's post count."},{"title":"Open saved evidence","credits":0,"includes":"Open a saved research link, or recover the same completed response through your agent."}],"basis":"These examples describe particular request steps, not fixed prices per answer. More authors, periods, sources or pages can require more data requests. A continuation delivers another page; exact recovery returns a response already paid for. Each response reports its actual charge.","free":["Connection and capability guidance","Current balance checks","Exact recovery of an already completed response","Opening your own saved research links","Website Live browsing"],"account_url":"https://cryptosocials.net/settings/account"},"data":{"title":"The evidence behind the answer","summary":"Selected Crypto X posts, linked to identified assets and dated price history, with original sources you can inspect.","sections":[{"title":"Selected sources","text":"Collection follows selected Crypto X accounts, with source checks that exclude unavailable, protected, suspended and blocked accounts from the active set. Quotes and reposts can also preserve material from other authors. Source identities and post relationships remain attached so you and your agent can examine who said what."},{"title":"Coin identities, not just tickers","text":"Posts are linked to specific assets using the identity evidence available. That matters when several coins share a symbol. Unresolved or ambiguous matches remain explicit, so an uncertain mention does not quietly become a confident coin match."},{"title":"Prices with dates and provenance","text":"Stored price observations and available candles can be read alongside posts. Source, effective time and capture time stay attached. That lets your agent compare a claim with what was observed afterward without pretending every post has an exact trade-time price."},{"title":"Live discussion and retained history","text":"Collection adds new evidence continuously while retaining earlier posts and price observations. Ask your agent for a recent window or look back at a past period. Freshness and history differ by source and asset; your agent can inspect the coverage relevant to your question."},{"title":"Sources you can check","text":"Research returns source links, identities and retained text. Available reply, quote and repost relationships help distinguish an original claim from a later share. Your agent interprets this evidence and cites what supports its answer."}]},"connection":{"setup_url":"https://cryptosocials.net/agent-access","chatgpt_plugin_url":"https://chatgpt.com/plugins/plugin_asdk_app_6ab1310a217c8191a407ab171972a4ed","steps":["Open CryptoSocials in your app’s Plugins section and select Install plugin or Connect.","Sign in to CryptoSocials, or create an account and verify your email, then approve read access.","Return to your app and ask your question using CryptoSocials."],"chatgpt_recovery":"If ChatGPT doesn't use CryptoSocials, open the plugin page, choose Try in chat, and paste your question.","account_help":"Install and approve the connection in the app before researching. Enter credentials only on CryptoSocials, never in chat.","manual_setup":"Manual Codex, other bearer-key agents and REST use an access key from Connections and keys."},"research_v2":{"contract_version":"2.0","interface":"accepted_nineteen_tools","guidance_version":"shared-research-guidance-20261021-v1","schema_url":"https://cryptosocials.net/developers/research-v2.json","tools":["discover_evidence","measure_evidence","read_evidence","read_price_context","get_research_capabilities","get_research_balance","list_research_requests","get_research_status","recover_research_result","cancel_research_request"],"rest_control":{"method":"POST","path":"/agent/v2/research/control","actions":["guide","balance","recent","status","recover","cancel"]},"legacy_workflows":"The topic-specific technical recipes document supported V1 calls. They apply only to connections advertising those tools; the published current MCP uses the named research tools and free controls listed here.","guidance":"Calculations cover the entire eligible stored population in the accepted dates and filters before delivery pagination. Original wording, available relationship context and paired source citations are distinct from numerical calculations. Complete the original review required by the question. Each result has its own scope, task and observation cutoff. Retained originals reproduce their captured cutoff; an intentionally fresh later reading does not refresh earlier figures. Temporary calculation results and retained evidence are available for 24 hours from the stated cutoff, until retention.expires_at. Follow-ups reuse that fixed evidence during the window. Expired references require explicitly fresh research with the complete requested scope; they are never silently refreshed or charged. The canonical post archive and billing receipts are separate. Successful data requests, including new delivery pages, use the shared CryptoSocials account tariff. Responses report the actual additional charge; routine balance polling is unnecessary. An exact retry or ready-result recovery returns the same saved response without an additional charge. It is not fresh research or another page. Unconfirmed accounting remains uncertain until recovered."}}