"Fossil SCM" matching MCP tools:

• openfinance_search_bank_connectorsA

  QI SCD MCPA

  Searches the available bank connectors by name (pass keywords[], e.g. ['nubank','btg']) and returns, per match: the connector id, whether it's Open Finance or API (`access`), PF/PJ (`audience`), the user's already-linked connections (and accounts when include_accounts=true), and a ready `connect_url` with the bank pre-selected. Some non-Open-Finance credential connectors carry a `caveat` warning that they don't auto-update (needs periodic manual reconnection) — surface it so the user can prefer the institution's Open Finance connector for automation. Honors the user's plan (a PF plan hides PJ banks; a PJ plan covers BOTH — PF and PJ banks connect and count under the same plan). Call this BEFORE connecting to hand the user a one-click link to the right bank. keywords[] is REQUIRED — without it returns a hint (never dumps the whole catalog).

  ConnectorNo auth

• openfinance_force_syncA

  QI SCD MCPA

  Forces the bank to re-sync one or more connections NOW and WAITS for it to finish (PATCH /items/:id, then polls until the item stops updating, up to ~60s). Use this when a balance or transaction list looks stale: a connection can read UPDATED yet be hours old, and this pulls fresh data WITHOUT disconnecting/reconnecting. Pass `items` as an array of selectors (item_id, connector_id, connector_name, or the user-set custom_label nickname); OMIT `items` to sync ALL linked banks. Returns `{ results, errors }`; each result has the final `status`, `executionStatus`, `lastUpdatedAt` (advances when data is refreshed), and `synced` (true = fresh data is ready). `needs_action` (e.g. MFA_REINTERACTION / LOGIN_ERROR / WAITING_USER_INPUT) means the user must re-authenticate — those results include a `reconnect_url` that opens the widget in UPDATE mode for that exact connection (user enters credentials / MFA token, data refreshes in place, no slot consumed, no disconnect needed). `timed_out: true` means the sync is still running — re-check with openfinance_get_item_status. Set `wait: false` for fire-and-forget (returns immediately while UPDATING).

  ConnectorNo auth

• openfinance_update_transaction_categoryA

  QI SCD MCPA

  Corrects the category of one or more transactions (PATCH /transactions/:id). Pass `items` as an array of { transaction_id, category_id } — `transaction_id` comes from openfinance_list_transactions, `category_id` from openfinance_list_categories. This overrides Pluggy's automatic categorization AND teaches Pluggy: recategorizing a transaction automatically creates a Category Rule for this client (case-insensitive exact match on the transaction's data), so FUTURE similar transactions are categorized the same way — use this to fix miscategorized transactions and improve categorization accuracy going forward. Batch shape: returns `{ updated, results: [{ transaction_id, category, categoryId }], errors: [{ id, status, message }] }` — per-item errors do not fail the whole batch.

  ConnectorNo auth

• polymarket_arbitrageA

  Herb TcmA

  Find arbitrage opportunities on Polymarket via monotonicity violations + partition-sum checks. Call with NO args for a `trending_scan` of the top ~200 markets by weekly volume; pass `event` for the strongest per-event partition_check, or `topic` for a themed cross-event scan. `event` (recommended for a specific market): pass a Polymarket event slug like "fed-decision-may-2026" or "when-will-bitcoin-hit-150k"; walks child markets, checks date-axis / threshold-axis ordering AND computes the partition_check (sum of YES prices across mutually-exclusive legs — should ≈1; deviations >3pp emit a BUY/SELL EVERY LEG signal). `topic` (for cross-event scanning): pass a seed question like "Strait of Hormuz traffic returns to normal" or "Fed rate decision"; searches related events across the platform, flattens markets, runs the comparator on the union. Cross-event mode catches "...by May 31" vs "...by Jun 30" patterns that single-event misses. SEMANTIC ANCHOR: cross-event pairs require ≥0.30 Jaccard similarity on question tokens (prevents Powell-Fed-Pause being paired with Powell-DOJ-probe); skipped_low_similarity surfaces the rejected pair count. PARTITION FILTER: drops will-person-X / will-manager-Y / will-someone-else- placeholder slugs; partitions with >20% placeholder fraction return null arb signal. Response: opportunities[] (gap_pp, suggested_trade, reasoning, monotonicity violation context), and in event mode partition_check{sum_yes_prices, gap_from_1, placeholders_filtered, suggested_trade}. FEES: every opportunities[] row and partition_check.arbitrage carry edge_pp_gross (== gap_pp / overround_pp), fees_pp, edge_pp_net, net_positive, plus polymarket_fee_pp, fee_basis and fee_categories[]. BOTH cost components are modeled: Polymarket's own per-category TAKER FEE (fee = shares × rate × p × (1-p), rates crypto 0.07 / sports-economics-culture-weather-other 0.05 / finance-politics-mentions-tech 0.04, geopolitics and world events fee-free; verified against Polymarket's own docs as of 2026-09-13) and Polygon gas (~$0.02/leg). The taker fee dominates: ~$1.75 per 100 shares on a crypto market at 50c versus $0.02 of gas, so rows that looked profitable before fleet #1927 may now show net_positive:false — that is the correction, not a regression. Each leg is priced at ITS OWN market's rate and price (the fee curve peaks at 50c and falls toward both extremes). fee_basis says where the rate came from: 'payload' (read off the market, the normal case), 'category' (mapped from its fee category), 'fee_free', or 'fallback' (rate unknown — charged at the modal 0.05 rather than assumed free, so an unreadable market is never reported as costless). Where fill_check reprices against live depth, this does NOT double-count that spread cost. FILL CHECK: when the partition signal fires, arbitrage.fill_check prices it against live CLOB depth (theoretical_edge_pp_at_book vs realizable_edge_pp at 1000 shares/leg, thin_legs[]) — realizable_edge_pp ≤ 0 means the overround exists only at last-trade, not in the book; do not trade it. For custom sizing use polymarket_fill_risk.

  ConnectorNo auth

• polymarket_arbitrageA

  Scb SeB

  Find arbitrage opportunities on Polymarket via monotonicity violations + partition-sum checks. Call with NO args for a `trending_scan` of the top ~200 markets by weekly volume; pass `event` for the strongest per-event partition_check, or `topic` for a themed cross-event scan. `event` (recommended for a specific market): pass a Polymarket event slug like "fed-decision-may-2026" or "when-will-bitcoin-hit-150k"; walks child markets, checks date-axis / threshold-axis ordering AND computes the partition_check (sum of YES prices across mutually-exclusive legs — should ≈1; deviations >3pp emit a BUY/SELL EVERY LEG signal). `topic` (for cross-event scanning): pass a seed question like "Strait of Hormuz traffic returns to normal" or "Fed rate decision"; searches related events across the platform, flattens markets, runs the comparator on the union. Cross-event mode catches "...by May 31" vs "...by Jun 30" patterns that single-event misses. SEMANTIC ANCHOR: cross-event pairs require ≥0.30 Jaccard similarity on question tokens (prevents Powell-Fed-Pause being paired with Powell-DOJ-probe); skipped_low_similarity surfaces the rejected pair count. PARTITION FILTER: drops will-person-X / will-manager-Y / will-someone-else- placeholder slugs; partitions with >20% placeholder fraction return null arb signal. Response: opportunities[] (gap_pp, suggested_trade, reasoning, monotonicity violation context), and in event mode partition_check{sum_yes_prices, gap_from_1, placeholders_filtered, suggested_trade}. FEES: every opportunities[] row and partition_check.arbitrage carry edge_pp_gross (== gap_pp / overround_pp), fees_pp, edge_pp_net, net_positive, plus polymarket_fee_pp, fee_basis and fee_categories[]. BOTH cost components are modeled: Polymarket's own per-category TAKER FEE (fee = shares × rate × p × (1-p), rates crypto 0.07 / sports-economics-culture-weather-other 0.05 / finance-politics-mentions-tech 0.04, geopolitics and world events fee-free; verified against Polymarket's own docs as of 2026-09-13) and Polygon gas (~$0.02/leg). The taker fee dominates: ~$1.75 per 100 shares on a crypto market at 50c versus $0.02 of gas, so rows that looked profitable before fleet #1927 may now show net_positive:false — that is the correction, not a regression. Each leg is priced at ITS OWN market's rate and price (the fee curve peaks at 50c and falls toward both extremes). fee_basis says where the rate came from: 'payload' (read off the market, the normal case), 'category' (mapped from its fee category), 'fee_free', or 'fallback' (rate unknown — charged at the modal 0.05 rather than assumed free, so an unreadable market is never reported as costless). Where fill_check reprices against live depth, this does NOT double-count that spread cost. FILL CHECK: when the partition signal fires, arbitrage.fill_check prices it against live CLOB depth (theoretical_edge_pp_at_book vs realizable_edge_pp at 1000 shares/leg, thin_legs[]) — realizable_edge_pp ≤ 0 means the overround exists only at last-trade, not in the book; do not trade it. For custom sizing use polymarket_fill_risk.

  ConnectorNo auth

• polymarket_arbitrageA

  Tcm ProductsA

  Find arbitrage opportunities on Polymarket via monotonicity violations + partition-sum checks. Call with NO args for a `trending_scan` of the top ~200 markets by weekly volume; pass `event` for the strongest per-event partition_check, or `topic` for a themed cross-event scan. `event` (recommended for a specific market): pass a Polymarket event slug like "fed-decision-may-2026" or "when-will-bitcoin-hit-150k"; walks child markets, checks date-axis / threshold-axis ordering AND computes the partition_check (sum of YES prices across mutually-exclusive legs — should ≈1; deviations >3pp emit a BUY/SELL EVERY LEG signal). `topic` (for cross-event scanning): pass a seed question like "Strait of Hormuz traffic returns to normal" or "Fed rate decision"; searches related events across the platform, flattens markets, runs the comparator on the union. Cross-event mode catches "...by May 31" vs "...by Jun 30" patterns that single-event misses. SEMANTIC ANCHOR: cross-event pairs require ≥0.30 Jaccard similarity on question tokens (prevents Powell-Fed-Pause being paired with Powell-DOJ-probe); skipped_low_similarity surfaces the rejected pair count. PARTITION FILTER: drops will-person-X / will-manager-Y / will-someone-else- placeholder slugs; partitions with >20% placeholder fraction return null arb signal. Response: opportunities[] (gap_pp, suggested_trade, reasoning, monotonicity violation context), and in event mode partition_check{sum_yes_prices, gap_from_1, placeholders_filtered, suggested_trade}. FEES: every opportunities[] row and partition_check.arbitrage carry edge_pp_gross (== gap_pp / overround_pp), fees_pp, edge_pp_net, net_positive, plus polymarket_fee_pp, fee_basis and fee_categories[]. BOTH cost components are modeled: Polymarket's own per-category TAKER FEE (fee = shares × rate × p × (1-p), rates crypto 0.07 / sports-economics-culture-weather-other 0.05 / finance-politics-mentions-tech 0.04, geopolitics and world events fee-free; verified against Polymarket's own docs as of 2026-09-13) and Polygon gas (~$0.02/leg). The taker fee dominates: ~$1.75 per 100 shares on a crypto market at 50c versus $0.02 of gas, so rows that looked profitable before fleet #1927 may now show net_positive:false — that is the correction, not a regression. Each leg is priced at ITS OWN market's rate and price (the fee curve peaks at 50c and falls toward both extremes). fee_basis says where the rate came from: 'payload' (read off the market, the normal case), 'category' (mapped from its fee category), 'fee_free', or 'fallback' (rate unknown — charged at the modal 0.05 rather than assumed free, so an unreadable market is never reported as costless). Where fill_check reprices against live depth, this does NOT double-count that spread cost. FILL CHECK: when the partition signal fires, arbitrage.fill_check prices it against live CLOB depth (theoretical_edge_pp_at_book vs realizable_edge_pp at 1000 shares/leg, thin_legs[]) — realizable_edge_pp ≤ 0 means the overround exists only at last-trade, not in the book; do not trade it. For custom sizing use polymarket_fill_risk.

  ConnectorNo auth

• openfinance_list_connectionsA

  QI SCD MCPA

  Returns the saved bank connections for this install: connector_id, item_id, bank name, a per-connection reconnect_url, and an add_connection_url to link additional banks via the Open Finance widget. The reconnect_url reopens the widget in UPDATE mode for that EXISTING connection (user re-enters credentials / MFA token and the data refreshes in place) — use it when a connection needs re-authentication (MFA connectors, LOGIN_ERROR, stale non-Open-Finance data). It does NOT consume a connection slot and does NOT require disconnecting first.

  ConnectorNo auth

• openfinance_provider_statusA

  QI SCD MCPA

  Checks the LIVE operational status of the Open Finance provider (its public status page) — this is the PROVIDER's health, separate from your own connection's `openfinance_get_item_status`. Use it whenever data looks incomplete or stale even though a connection shows UPDATED (accounts/transactions/balances missing, a bank not returning everything): it reveals an upstream outage or a known incident on a specific bank/connector, so you can tell a provider-side problem apart from a connection that just needs reconnecting. Returns the global indicator (none/minor/major/critical), degraded components, open incidents, and — when you have banks connected — flags the incidents that affect YOUR connected banks in `your_banks_affected`.

  ConnectorNo auth

• paleobiology_dataframe_queryA

  paleobiology-mcp-serverA

  Run a read-only SQL SELECT against occurrence result sets staged on a DataCanvas by paleobiology_search_occurrences. This is how you analyze a large fossil set without re-fetching it: count occurrences by early_interval, group by formation, country (cc), or accepted_name, or filter by a paleo/modern coordinate range. The classification column is JSON — roll up by rank with json_extract_string(classification, '$.family') (also $.phylum, $.class, $.order, $.genus). Staged rows are occurrences, so collection-only fields such as lithology are not present. Reference tables by the table_name that paleobiology_search_occurrences returned — call paleobiology_dataframe_describe first if you do not know the table or column names. Every row supplied by the canvas is returned; choose smaller pages with SQL LIMIT/OFFSET. Results exceeding the canvas row cap include a notice naming the stable ORDER BY … LIMIT/OFFSET continuation. SELECT only; writes and file-reading functions are rejected.

  ConnectorNo auth

• kalshi_weather_edgeA

  Herb TcmA

  Prices Kalshi daily high-temperature markets against the NWS forecast for the market's OWN settlement station, and measures whether that forecast actually beats the market. Two modes. LIVE (default): returns the full strike ladder for one city and settlement date with market_prob (mid), forecast_prob, and edge_pp per strike, plus the settlement clause verbatim. BACKTEST (`backtest_days: N`): scores an archived gridded forecast against the market on settled days and returns brier_market vs brier_forecast with a plain-English `verdict`, so the edge is MEASURED rather than asserted. READ THE WARNINGS — they are not boilerplate. (1) These markets DO NOT settle on the NWS. They settle on The Weather Company (weather.com) at a Kalshi station code such as CLINYC, which the response quotes verbatim; so part of every edge_pp is NWS-vs-Weather-Company disagreement about the same day at the same station, which is not mispricing and not tradeable. `settlement_vs_forecast_basis_f` from backtest mode is that part as a number. (2) The station is DERIVED from the settlement clause, never from the city name: Chicago settles at MIDWAY and New York at CENTRAL PARK, so a city-centre forecast would misprice a whole ladder. A station that cannot be resolved yields rows with no forecast and a reason, never a guessed coordinate. (3) forecast_prob assumes a normal distribution around the NWS high whose width is ASSUMED, not fitted (stated in `distribution_assumption`) — run backtest mode to see whether it is calibrated. (4) edge_pp is gross: no Kalshi fees, no bid-ask. MEASURED RESULT, AND IT IS NOT THE FLATTERING ONE: on the first backtest (KXHIGHNY, 13 settled days to 2026-09-11, 58 market observations) the MARKET beat the forecast — Brier 0.1008 for the market against 0.1594 for the archived gridded forecast, lower being better. So on that sample there is NO forecast edge to sell, and a large edge_pp is more likely to be the model disagreeing with a better-informed market than an opportunity. The measured settlement-vs-forecast basis was 1.7F mean absolute over 8 pinnable days, slightly warm-biased, which is a big share of a typical edge_pp on a 2-degree bracket. Re-run backtest_days before believing any edge; if a later sample reverses this, the numbers say so. NWS is US-only, so the ~30 international Kalshi weather series (London, Paris, Tokyo) return market prices with forecast_unavailable rather than a forecast. Precipitation series are listed but not yet priced. Cities: nyc, chicago, los angeles, miami, austin, houston, denver, philadelphia — or pass `series_ticker` for any other (e.g. "KXHIGHTBOS").

  ConnectorNo auth

• research_rowA

  Herb TcmA

  Claygent-style per-row research: answer ONE question about ONE company and return a JSON object matching your output_schema, with every field wrapped in its own evidence — {value, evidence_url, quote, status}. status is NEVER self-reported by the model: "found" means this code independently verified the model's quote is a VERBATIM substring of text actually fetched; "inferred" means a value was given but the quote could not be verified; "not_established" means no source answered the field, and its value is always null — never guessed, never a blended score. Sources: "web" fetches the company's own site (entity.domain required — homepage, /about, /leadership) via fetch_url; "catalog" checks entity_profile (SEC EDGAR; resolved:false for a private company is expected, not a failure); "both" (default) tries both. max_cost (USD, default 0.02, max 0.50) scales how many pages are fetched and refuses up front if it is too small to pay for even one extraction call (floor $0.002) — never silently overspends. A row where every source comes back empty costs nothing: every field is mechanically not_established with NO model call made. For a boolean/derived field (e.g. "does this company sell to healthcare"), the quote only has to SUPPORT the value — a descriptive sentence about customers/products counts, the source need not state the answer verbatim. leadership_page_status ("fetched"|"not_found"|"not_attempted", plus leadership_note) says plainly whether a leadership/team page was actually read, so a named-person field (head_of_sales) coming back not_established is distinguishable from "this company just doesn't publish one" vs. "we never checked". model_output_empty:true means the model returned nothing usable (a transport/budget problem) — every field is not_established for THAT reason, never read as 20 individual "no" judgments. Use with a fixed question + schema across many rows for "Claygent over N companies with a hard cost cap" (composes with batch callers). For picking from a fixed label list instead of extracting open fields, use classify_row (cheaper, one model call).

  ConnectorNo auth

• pipeworx_feedbackA

  Herb TcmA

  Tell the Pipeworx team something is broken, missing, or needs to exist. Use when a tool returns wrong/stale data (bug), when a tool you wish existed isn't in the catalog (feature/data_gap), or when something worked surprisingly well (praise). ONLY for tools served by this Pipeworx connection — if the tool came from a different MCP server in your client (another vendor's Gmail, Splunk, Slack, etc. connector), we cannot fix it and reporting it here only delays you; file it with that server instead. Not sure? Pipeworx tool names are the ones this connection lists. Describe the issue in terms of Pipeworx tools/packs — don't paste the end-user's prompt. Filing without an account returns a `claim_token`; pass it back later as pipeworx_feedback({claim_token:"pwfb_…"}) to read whether it was fixed and what changed. The team reads digests daily and signal directly affects roadmap. Rate-limited to 5 per identifier per day. Free; doesn't count against your tool-call quota.

  ConnectorNo auth

• kalshi_weather_edgeA

  Scb SeB

  Prices Kalshi daily high-temperature markets against the NWS forecast for the market's OWN settlement station, and measures whether that forecast actually beats the market. Two modes. LIVE (default): returns the full strike ladder for one city and settlement date with market_prob (mid), forecast_prob, and edge_pp per strike, plus the settlement clause verbatim. BACKTEST (`backtest_days: N`): scores an archived gridded forecast against the market on settled days and returns brier_market vs brier_forecast with a plain-English `verdict`, so the edge is MEASURED rather than asserted. READ THE WARNINGS — they are not boilerplate. (1) These markets DO NOT settle on the NWS. They settle on The Weather Company (weather.com) at a Kalshi station code such as CLINYC, which the response quotes verbatim; so part of every edge_pp is NWS-vs-Weather-Company disagreement about the same day at the same station, which is not mispricing and not tradeable. `settlement_vs_forecast_basis_f` from backtest mode is that part as a number. (2) The station is DERIVED from the settlement clause, never from the city name: Chicago settles at MIDWAY and New York at CENTRAL PARK, so a city-centre forecast would misprice a whole ladder. A station that cannot be resolved yields rows with no forecast and a reason, never a guessed coordinate. (3) forecast_prob assumes a normal distribution around the NWS high whose width is ASSUMED, not fitted (stated in `distribution_assumption`) — run backtest mode to see whether it is calibrated. (4) edge_pp is gross: no Kalshi fees, no bid-ask. MEASURED RESULT, AND IT IS NOT THE FLATTERING ONE: on the first backtest (KXHIGHNY, 13 settled days to 2026-09-11, 58 market observations) the MARKET beat the forecast — Brier 0.1008 for the market against 0.1594 for the archived gridded forecast, lower being better. So on that sample there is NO forecast edge to sell, and a large edge_pp is more likely to be the model disagreeing with a better-informed market than an opportunity. The measured settlement-vs-forecast basis was 1.7F mean absolute over 8 pinnable days, slightly warm-biased, which is a big share of a typical edge_pp on a 2-degree bracket. Re-run backtest_days before believing any edge; if a later sample reverses this, the numbers say so. NWS is US-only, so the ~30 international Kalshi weather series (London, Paris, Tokyo) return market prices with forecast_unavailable rather than a forecast. Precipitation series are listed but not yet priced. Cities: nyc, chicago, los angeles, miami, austin, houston, denver, philadelphia — or pass `series_ticker` for any other (e.g. "KXHIGHTBOS").

  ConnectorNo auth

• research_rowA

  Scb SeB

  Claygent-style per-row research: answer ONE question about ONE company and return a JSON object matching your output_schema, with every field wrapped in its own evidence — {value, evidence_url, quote, status}. status is NEVER self-reported by the model: "found" means this code independently verified the model's quote is a VERBATIM substring of text actually fetched; "inferred" means a value was given but the quote could not be verified; "not_established" means no source answered the field, and its value is always null — never guessed, never a blended score. Sources: "web" fetches the company's own site (entity.domain required — homepage, /about, /leadership) via fetch_url; "catalog" checks entity_profile (SEC EDGAR; resolved:false for a private company is expected, not a failure); "both" (default) tries both. max_cost (USD, default 0.02, max 0.50) scales how many pages are fetched and refuses up front if it is too small to pay for even one extraction call (floor $0.002) — never silently overspends. A row where every source comes back empty costs nothing: every field is mechanically not_established with NO model call made. For a boolean/derived field (e.g. "does this company sell to healthcare"), the quote only has to SUPPORT the value — a descriptive sentence about customers/products counts, the source need not state the answer verbatim. leadership_page_status ("fetched"|"not_found"|"not_attempted", plus leadership_note) says plainly whether a leadership/team page was actually read, so a named-person field (head_of_sales) coming back not_established is distinguishable from "this company just doesn't publish one" vs. "we never checked". model_output_empty:true means the model returned nothing usable (a transport/budget problem) — every field is not_established for THAT reason, never read as 20 individual "no" judgments. Use with a fixed question + schema across many rows for "Claygent over N companies with a hard cost cap" (composes with batch callers). For picking from a fixed label list instead of extracting open fields, use classify_row (cheaper, one model call).

  ConnectorNo auth

• pipeworx_feedbackA

  Scb SeB

  Tell the Pipeworx team something is broken, missing, or needs to exist. Use when a tool returns wrong/stale data (bug), when a tool you wish existed isn't in the catalog (feature/data_gap), or when something worked surprisingly well (praise). ONLY for tools served by this Pipeworx connection — if the tool came from a different MCP server in your client (another vendor's Gmail, Splunk, Slack, etc. connector), we cannot fix it and reporting it here only delays you; file it with that server instead. Not sure? Pipeworx tool names are the ones this connection lists. Describe the issue in terms of Pipeworx tools/packs — don't paste the end-user's prompt. Filing without an account returns a `claim_token`; pass it back later as pipeworx_feedback({claim_token:"pwfb_…"}) to read whether it was fixed and what changed. The team reads digests daily and signal directly affects roadmap. Rate-limited to 5 per identifier per day. Free; doesn't count against your tool-call quota.

  ConnectorNo auth

• kalshi_weather_edgeA

  Tcm ProductsA

  Prices Kalshi daily high-temperature markets against the NWS forecast for the market's OWN settlement station, and measures whether that forecast actually beats the market. Two modes. LIVE (default): returns the full strike ladder for one city and settlement date with market_prob (mid), forecast_prob, and edge_pp per strike, plus the settlement clause verbatim. BACKTEST (`backtest_days: N`): scores an archived gridded forecast against the market on settled days and returns brier_market vs brier_forecast with a plain-English `verdict`, so the edge is MEASURED rather than asserted. READ THE WARNINGS — they are not boilerplate. (1) These markets DO NOT settle on the NWS. They settle on The Weather Company (weather.com) at a Kalshi station code such as CLINYC, which the response quotes verbatim; so part of every edge_pp is NWS-vs-Weather-Company disagreement about the same day at the same station, which is not mispricing and not tradeable. `settlement_vs_forecast_basis_f` from backtest mode is that part as a number. (2) The station is DERIVED from the settlement clause, never from the city name: Chicago settles at MIDWAY and New York at CENTRAL PARK, so a city-centre forecast would misprice a whole ladder. A station that cannot be resolved yields rows with no forecast and a reason, never a guessed coordinate. (3) forecast_prob assumes a normal distribution around the NWS high whose width is ASSUMED, not fitted (stated in `distribution_assumption`) — run backtest mode to see whether it is calibrated. (4) edge_pp is gross: no Kalshi fees, no bid-ask. MEASURED RESULT, AND IT IS NOT THE FLATTERING ONE: on the first backtest (KXHIGHNY, 13 settled days to 2026-09-11, 58 market observations) the MARKET beat the forecast — Brier 0.1008 for the market against 0.1594 for the archived gridded forecast, lower being better. So on that sample there is NO forecast edge to sell, and a large edge_pp is more likely to be the model disagreeing with a better-informed market than an opportunity. The measured settlement-vs-forecast basis was 1.7F mean absolute over 8 pinnable days, slightly warm-biased, which is a big share of a typical edge_pp on a 2-degree bracket. Re-run backtest_days before believing any edge; if a later sample reverses this, the numbers say so. NWS is US-only, so the ~30 international Kalshi weather series (London, Paris, Tokyo) return market prices with forecast_unavailable rather than a forecast. Precipitation series are listed but not yet priced. Cities: nyc, chicago, los angeles, miami, austin, houston, denver, philadelphia — or pass `series_ticker` for any other (e.g. "KXHIGHTBOS").

  ConnectorNo auth

• research_rowA

  Tcm ProductsA

  Claygent-style per-row research: answer ONE question about ONE company and return a JSON object matching your output_schema, with every field wrapped in its own evidence — {value, evidence_url, quote, status}. status is NEVER self-reported by the model: "found" means this code independently verified the model's quote is a VERBATIM substring of text actually fetched; "inferred" means a value was given but the quote could not be verified; "not_established" means no source answered the field, and its value is always null — never guessed, never a blended score. Sources: "web" fetches the company's own site (entity.domain required — homepage, /about, /leadership) via fetch_url; "catalog" checks entity_profile (SEC EDGAR; resolved:false for a private company is expected, not a failure); "both" (default) tries both. max_cost (USD, default 0.02, max 0.50) scales how many pages are fetched and refuses up front if it is too small to pay for even one extraction call (floor $0.002) — never silently overspends. A row where every source comes back empty costs nothing: every field is mechanically not_established with NO model call made. For a boolean/derived field (e.g. "does this company sell to healthcare"), the quote only has to SUPPORT the value — a descriptive sentence about customers/products counts, the source need not state the answer verbatim. leadership_page_status ("fetched"|"not_found"|"not_attempted", plus leadership_note) says plainly whether a leadership/team page was actually read, so a named-person field (head_of_sales) coming back not_established is distinguishable from "this company just doesn't publish one" vs. "we never checked". model_output_empty:true means the model returned nothing usable (a transport/budget problem) — every field is not_established for THAT reason, never read as 20 individual "no" judgments. Use with a fixed question + schema across many rows for "Claygent over N companies with a hard cost cap" (composes with batch callers). For picking from a fixed label list instead of extracting open fields, use classify_row (cheaper, one model call).

  ConnectorNo auth

• pipeworx_feedbackA

  Tcm ProductsA

  Tell the Pipeworx team something is broken, missing, or needs to exist. Use when a tool returns wrong/stale data (bug), when a tool you wish existed isn't in the catalog (feature/data_gap), or when something worked surprisingly well (praise). ONLY for tools served by this Pipeworx connection — if the tool came from a different MCP server in your client (another vendor's Gmail, Splunk, Slack, etc. connector), we cannot fix it and reporting it here only delays you; file it with that server instead. Not sure? Pipeworx tool names are the ones this connection lists. Describe the issue in terms of Pipeworx tools/packs — don't paste the end-user's prompt. Filing without an account returns a `claim_token`; pass it back later as pipeworx_feedback({claim_token:"pwfb_…"}) to read whether it was fixed and what changed. The team reads digests daily and signal directly affects roadmap. Rate-limited to 5 per identifier per day. Free; doesn't count against your tool-call quota.

  ConnectorNo auth

• openfinance_list_accountsA

  QI SCD MCPA

  Returns accounts for a bank connection: BANK (checking/savings) and CREDIT (credit card) with balance, number, type, subtype, bankData, and creditData. Also returns `bank` (the brand/connector name like 'Nubank Empresas' — same shown in the dashboard UI) and `connector_id`. Note: each account's `name` is the legal entity that issues the account (e.g. 'Nu Pagamentos S.A. - Instituição de Pagamento'), which is not the same as the brand — when referring to the bank in user-facing text, use `bank`. OMIT `item` to list accounts across ALL linked banks at once — the response aggregates every connection's accounts into `results`, each row tagged with its own `bank`/`connector_id`/`item_id` (use this when the user asks for 'my accounts/cards' without naming a bank). Pass `item` to target a single bank (response carries `bank`/`connector_id`/`item_id` at the root). CREDIT (credit card) `balance`: its meaning is CONNECTOR-DEPENDENT — some banks report the current open-bill partial, others the full revolving/installment debt — so do NOT treat `balance` as 'this month's bill'. The open billing cycle is defined by `creditData.balanceCloseDate` (when it closes) / `balanceDueDate` (when it's due). For a standardized open-bill amount and total debt that mean the same across connectors, use openfinance_list_credit_card_bills (`open_bill` + `total_pending_debt`, derived from PENDING transactions); closed bills come from that same tool's `results`. A CREDIT row may carry `creditData.usedAmount` (how much of THIS card's limit the bank reports as consumed) and a `balance_notice`. `balance_notice` means `balance` came back 0,00 while the bank's own payload indicates an outstanding amount — some issuers never fill the card's consolidated balance field. When it is present, do NOT tell the user the card has nothing to pay: read the amount from openfinance_list_credit_card_bills instead. `bankData.closingBalance` and `automaticallyInvestedBalance` are provider-reported extras that can LAG right after a connection is first created: the bank may publish the connection as UPDATED before those derived fields converge, so they can briefly carry a stale/phantom value that a force sync (openfinance_force_sync) reconciles. The account's own `balance` is authoritative — treat those two as hints until they agree with it. May include a `provider_incident` block when the Open Finance provider has an OPEN incident affecting a bank in this response: balances and credit limits may be unreliable (incomplete or wrong, e.g. a credit limit near 1,00) even with the connection UPDATED, until the provider recovers. Do not present those values as real. May include an `identity_notice` when the SAME account (same number) arrives via two connections stamped with DIFFERENT owner/taxNumber: in Open Finance those fields reflect each connection's CONSENT HOLDER (e.g. a joint account consented by both holders), so dedupe by account number before summing balances and do not attribute ownership by owner/taxNumber for those accounts. Bulk support: accepts item_ids for batched execution.

  ConnectorNo auth

• openfinance_list_transactionsA

  QI SCD MCPA

  Returns transactions for a bank account (BANK or CREDIT type). For CREDIT (credit card) accounts, this is the ONLY way to get itemized transactions (purchases, subscriptions, etc.). Each credit card transaction MAY carry `creditCardMetadata.billId` pointing at a bill from openfinance_list_credit_card_bills, but this is a per-connector HINT, not authoritative: some connectors (e.g. Nubank) populate it sparsely (many transactions and installments arrive with no billId) or inconsistently (the same payment tagged to more than one bill). Do NOT reconstruct a bill's total by summing transactions by billId — the bill's own `totalAmount` from openfinance_list_credit_card_bills is the source of truth. CREDIT PENDING vs POSTED varies by connector: where the bank exposes future-dated `status:'PENDING'` installments, those represent the OPEN bill plus future bills (future months); where it does NOT, only the last closed bill's POSTED items appear until ~closing. Same query, different coverage per bank (upstream). To get a standardized open-bill total / total debt regardless, use openfinance_list_credit_card_bills (`open_bill` / `total_pending_debt`). SCHEDULED (future-dated) ROWS: results are ordered by date DESCENDING, and on a card with long installment plans the TOP of the list is the FUTURE — rows dated months ahead are scheduled installments of purchases already made, not new purchases. Every such row is flagged `scheduled:true`, the response carries `scheduled_count` and a `notice` naming the most recent row that actually happened. NEVER read the first row as 'the latest purchase' without checking `scheduled`. To list only what already happened, pass `to` = today. Supports from/to date filters (ISO YYYY-MM-DD) and an optional keyword filter via `search_queries` (case- and accent-insensitive substring match against description and merchant name, OR semantics across multiple terms). When `search_queries` is set the tool aggregates up to 5000 transactions within from/to before filtering — narrow from/to if `truncated:true` is returned. PAGINATION: OMIT both `page` and `page_size` (the default) to get ALL transactions in the from/to range in one call — the tool auto-paginates the upstream and returns them under a single logical page (`page:1`, `totalPages:1`), up to a 5000 ceiling (`truncated:true` + warning if exceeded, then narrow from/to). Passing `page` and/or `page_size` switches to MANUAL pagination: you get one page (`page_size` items, default 50, max 500; `page` defaults to 1) with the REAL `total`/`totalPages`, so `page_size:5` alone returns the first 5 with `totalPages` telling you how many pages remain. If the provider is temporarily unavailable or rate-limited, the call FAILS with an error saying when to retry (nothing was read, it does not mean there are no transactions); an unknown `account_id` fails with the valid accounts listed. With `search_queries`, a scan interrupted midway returns the matches found so far flagged `complete:false` + `incomplete_reason`. `detail` controls how much per-row data you get (default `'compact'` = slim, cheap). Use `detail:'rich'` to enrich each row (when the bank connector provides it) with `merchantInfo` (estabelecimento: businessName/razão social, cnpj, cnae, category — useful for auto-classifying spending) and extra `creditCardMetadata` fields: `billId` (a per-connector HINT toward the transaction's bill — sparse/inconsistent on some connectors like Nubank, so do NOT sum by it to get a bill total; use the bill's `totalAmount` instead), `billForecastDate`, `cardNumber`, `purchaseDate`, `payeeMCC`, `feeType`/`feeTypeAdditionalInfo`, `otherCreditsType`/`otherCreditsAdditionalInfo`. `billForecastDate` ("YYYY-MM") is the counterpart of `billId` for the OPEN cycle: PENDING transactions have NO billId (the bank only mints it once the bill closes), so this is the only field telling you which bill a pending purchase will land in — its month OFFSET is per-connector (some banks month+1, others month+0), so surface it as-is and do not derive a due date from it. `cardNumber` (last 4 digits) separates the primary cardholder's charges from an additional card's when several cards share one account_id. Use `detail:'raw'` to get the FULL untouched Pluggy transaction object (everything Pluggy returns, un-normalized — heaviest, for when you need a field we don't project). 'rich'/'raw' add tokens per row and coverage varies by bank/Open Finance, so keep the default for normal listings. For the card's statement closing/due dates use openfinance_list_accounts (`creditData.balanceCloseDate` / `balanceDueDate`). The response opens with an `account` echo block ({ account_id, bank, name, number, type, item_id }) identifying WHICH account/bank these transactions belong to. When more than one bank is connected, ALWAYS cross-check the echo against the account you intended to query and name the bank when presenting results — never attribute one bank's transactions to another. If total is 0 for a CREDIT account, check the connection health via openfinance_get_item_status — `statusDetail.creditCards.isUpdated: false` means the credit card sync failed and a force sync (openfinance_force_sync) or reconnection may be needed. May include a `provider_incident` block when the Open Finance provider has an OPEN incident affecting a connected bank: transactions may come back incomplete or wrong until the provider recovers, and reconnecting does not fix it. Bulk support: accepts account_ids for batched execution.

  ConnectorNo auth

• openfinance_list_loansA

  QI SCD MCPA

  Lists loan contracts per bank connection (GET /loans). Pass `items` as an array of connection selectors (item_id uuid, connector_id, or connector_name) — one entry per connection to fetch; multiple connections are queried sequentially with rate-limit spacing. OMIT `items` to list loans across ALL linked banks. Returns `{ results, errors }` per connection.

  ConnectorNo auth