{ "note": "Verbatim stdio transcript: OfferSheet MCP server spawned from a FRESH CLONE of remote main (not the working tree), keyless. Count is read from the tools/list response, not from any spec table.", "serverPath": "/tmp/offersheet-gpush/backend/mcp-server.js", "requests": [ { "jsonrpc": "2.0", "id": 1, "method": "initialize", "params": { "protocolVersion": "2024-11-05", "capabilities": {}, "clientInfo": { "name": "fresh-clone-probe", "version": "0" } } }, { "jsonrpc": "2.0", "method": "notifications/initialized" }, { "jsonrpc": "2.0", "id": 2, "method": "tools/list" } ], "responses": [ { "jsonrpc": "2.0", "id": 1, "result": { "protocolVersion": "2024-11-05", "capabilities": { "tools": {} }, "serverInfo": { "name": "offersheet", "version": "0.1.0" } } }, { "jsonrpc": "2.0", "id": 2, "result": { "tools": [ { "name": "offersheet_verdict_run", "description": "Runs the full OfferSheet verdict (Qloo evidence + Ticketmaster room + money + arms). Does NOT invent numbers: a withheld, price-unknown, or needs-input verdict is a success response with that status, never a number. Spends Qloo quota (guarded).", "inputSchema": { "type": "object", "properties": { "artist": { "type": "string" }, "city": { "type": "string" }, "guarantee": { "type": "number", "minimum": 0 }, "ticketPrice": { "type": "number", "exclusiveMinimum": 0 }, "variableCost": { "type": "number", "minimum": 0 }, "capacity": { "type": "number", "exclusiveMinimum": 0 }, "venue": { "type": "string" }, "entityHint": { "type": "string" } }, "required": [ "artist", "city", "guarantee" ], "additionalProperties": false } }, { "name": "offersheet_verdict_race", "description": "Same verdict as offersheet_verdict_run, plus arm timings (race: gutMs, webMs, totalMs). Does NOT change the verdict; it proves arm timing, which the plain verdict does not report. Spends Qloo quota (guarded).", "inputSchema": { "type": "object", "properties": { "artist": { "type": "string" }, "city": { "type": "string" }, "guarantee": { "type": "number", "minimum": 0 }, "ticketPrice": { "type": "number", "exclusiveMinimum": 0 }, "variableCost": { "type": "number", "minimum": 0 }, "capacity": { "type": "number", "exclusiveMinimum": 0 }, "venue": { "type": "string" }, "entityHint": { "type": "string" } }, "required": [ "artist", "city", "guarantee" ], "additionalProperties": false } }, { "name": "qloo_capabilities", "description": "Guarded proxy of Qloo's own qloo_capabilities tool (the pipeline readiness probe). Returns the upstream envelope plus an OfferSheet receipt. Does NOT invent a capability list when the harness is down — that errors. Spends Qloo quota (guarded).", "inputSchema": { "type": "object", "properties": { "operation_ids": { "type": "array", "items": { "type": "string" } } }, "required": [], "additionalProperties": false } }, { "name": "qloo_find_tags", "description": "Guarded proxy of Qloo's own qloo_find_tags tool. Tag URNs returned here feed the REST insights leg. Does NOT resolve an artist by itself. Spends Qloo quota (guarded).", "inputSchema": { "type": "object", "properties": { "query": { "type": "string" }, "limit": { "type": "number", "minimum": 1 } }, "required": [ "query" ], "additionalProperties": false } }, { "name": "qloo_describe", "description": "Guarded proxy of Qloo's own qloo_describe tool. An ambiguous name returns needs_input with candidates — it is never resolved by guessing. Spends Qloo quota (guarded).", "inputSchema": { "type": "object", "properties": { "entity": { "type": "string" }, "type": { "type": "string" } }, "required": [ "entity" ], "additionalProperties": false } }, { "name": "qloo_recommend", "description": "Guarded proxy of Qloo's own qloo_recommend tool (opener candidates for a headliner signal). Does NOT rank the candidates — that is qloo_rank, or the sequenced offersheet_opener_shortlist. Spends Qloo quota (guarded).", "inputSchema": { "type": "object", "properties": { "target_type": { "type": "string" }, "signals": { "type": "array", "items": { "type": "string" } }, "limit": { "type": "number", "minimum": 1 } }, "required": [ "target_type", "signals" ], "additionalProperties": false } }, { "name": "qloo_rank", "description": "Guarded proxy of Qloo's own qloo_rank tool. It ranks options a caller supplies; it does NOT generate candidates. Spends Qloo quota (guarded).", "inputSchema": { "type": "object", "properties": { "options": { "type": "array", "items": { "type": "string" } }, "option_type": { "type": "string" }, "signals": { "type": "array", "items": { "type": "string" } }, "signal_location": { "type": "string" } }, "required": [ "options", "option_type", "signals" ], "additionalProperties": false } }, { "name": "qloo_where_popular", "description": "Guarded proxy of Qloo's own qloo_where_popular tool. Returns matched areas/entities only — Qloo returns no headcount, and this tool does NOT invent one. Spends Qloo quota (guarded).", "inputSchema": { "type": "object", "properties": { "entity": { "type": "string" }, "within": { "type": "string" }, "limit": { "type": "number", "minimum": 1 } }, "required": [ "entity" ], "additionalProperties": false } }, { "name": "qloo_audience_demographics", "description": "Guarded proxy of Qloo's own qloo_audience_demographics tool. Bands carry no entity IDs, so their receipts honestly carry no ID list. Spends Qloo quota (guarded).", "inputSchema": { "type": "object", "properties": { "entity": { "type": "string" }, "entity_type": { "type": "string" } }, "required": [ "entity" ], "additionalProperties": false } }, { "name": "qloo_entity_tags", "description": "Guarded proxy of Qloo's own qloo_entity_tags tool (shared tags for a set of entities). One of the two accepted overlap evidences, alongside qloo_compare_audiences. Spends Qloo quota (guarded).", "inputSchema": { "type": "object", "properties": { "entities": { "type": "array", "items": { "type": "string" } }, "entity_type": { "type": "string" }, "limit": { "type": "number", "minimum": 1 } }, "required": [ "entities" ], "additionalProperties": false } }, { "name": "qloo_trends", "description": "Guarded proxy of Qloo's own qloo_trends tool (momentum over a date window). A failure is a labeled omission in a verdict, never a verdict blocker. Spends Qloo quota (guarded).", "inputSchema": { "type": "object", "properties": { "entities": { "type": "array", "items": { "type": "string" } }, "entity_type": { "type": "string" }, "start_date": { "type": "string" }, "end_date": { "type": "string" } }, "required": [ "entities" ], "additionalProperties": false } }, { "name": "qloo_compare_audiences", "description": "Guarded proxy of Qloo's own qloo_compare_audiences tool (audience-overlap object for two groups). Distinct from qloo_entity_tags (shared-tag list) and from offersheet_rest_compare (REST transport). Spends Qloo quota (guarded).", "inputSchema": { "type": "object", "properties": { "group_a": { "type": "array", "items": { "type": "string" } }, "group_b": { "type": "array", "items": { "type": "string" } }, "target_type": { "type": "string" } }, "required": [ "group_a", "group_b", "target_type" ], "additionalProperties": false } }, { "name": "offersheet_rest_search", "description": "REST name resolution via GET /search (key in the X-Api-Key header only, never in the URL). Accepts entities- or results-shaped bodies. Does NOT spend an MCP call. Spends Qloo quota (guarded).", "inputSchema": { "type": "object", "properties": { "query": { "type": "string" }, "take": { "type": "number", "minimum": 1 } }, "required": [ "query" ], "additionalProperties": false } }, { "name": "offersheet_rest_insights_by_tags", "description": "REST insights by tag URNs via GET /v2/insights with explainability on. Runs only when tag URNs are supplied; otherwise the caller reports it skipped. Spends Qloo quota (guarded).", "inputSchema": { "type": "object", "properties": { "tagUrns": { "type": "array", "items": { "type": "string" } }, "filterType": { "type": "string" } }, "required": [ "tagUrns" ], "additionalProperties": false } }, { "name": "offersheet_rest_compare", "description": "REST headliner-vs-opener comparison via GET /v2/analysis/compare. An object body with non-error keys counts as a payload. Spends Qloo quota (guarded).", "inputSchema": { "type": "object", "properties": { "aIds": { "type": "array", "items": { "type": "string" } }, "bIds": { "type": "array", "items": { "type": "string" } } }, "required": [ "aIds", "bIds" ], "additionalProperties": false } }, { "name": "offersheet_rest_entities", "description": "REST entity hydration via GET /entities. At most 15 IDs are sent per call (enforced, not advisory) — extra IDs are not sent. Spends Qloo quota (guarded).", "inputSchema": { "type": "object", "properties": { "ids": { "type": "array", "items": { "type": "string" } } }, "required": [ "ids" ], "additionalProperties": false } }, { "name": "offersheet_receipt_replay", "description": "Re-runs a receipted call live (MCP allowlist or REST allowlist) and compares fresh ordered entity IDs against originalIds (match / mismatch / not_comparable). The Qloo key gate precedes validation: a keyless bogus replay returns NOT_CONFIGURED before any other check. Demographics/overlap receipts carry no IDs, so they honestly report not_comparable. Spends Qloo quota (guarded).", "inputSchema": { "type": "object", "properties": { "kind": { "type": "string", "enum": [ "mcp", "rest" ] }, "tool": { "type": "string" }, "args": { "type": "object" }, "path": { "type": "string" }, "params": { "type": "object" }, "originalIds": { "type": "array", "items": { "type": "string" } } }, "required": [ "kind" ], "additionalProperties": false } }, { "name": "offersheet_receipt_verify_hash", "description": "Recomputes a receipt's SHA-256 offline (MCP: tool + args; REST: method + url) and reports valid true/false plus the recomputed hash. Does NOT make a live call and does NOT prove the upstream data — only that the hash matches the claimed request. Spends no Qloo calls.", "inputSchema": { "type": "object", "properties": { "operation": { "type": "string" }, "requestHash": { "type": "string" }, "tool": { "type": "string" }, "args": { "type": "object" }, "method": { "type": "string" }, "url": { "type": "string" } }, "required": [ "operation", "requestHash" ], "additionalProperties": false } }, { "name": "offersheet_receipts_list", "description": "Lists receipts from the session store (this process only; restart clears it) or reads the repo manifest (read-only; currently empty, returned verbatim with its note). The id / requestHash / operation parameters ARE the get operation — there is no separate get tool. This server never writes either source. Spends no Qloo calls.", "inputSchema": { "type": "object", "properties": { "source": { "type": "string", "enum": [ "session", "manifest" ] }, "id": { "type": "string" }, "requestHash": { "type": "string" }, "operation": { "type": "string" } }, "required": [], "additionalProperties": false } }, { "name": "offersheet_trials_list", "description": "Lists recorded benchmark trials from bench/results.json, labeled recorded, not live on every response. The index parameter IS the get operation. A missing file returns the honest no-runs note, never fabricated runs. Spends no Qloo calls.", "inputSchema": { "type": "object", "properties": { "index": { "type": "number", "minimum": 0 } }, "required": [], "additionalProperties": false } }, { "name": "offersheet_trial_pack_verify", "description": "Offline verification of frontend/trial-pack.json against bench/results.json: per-trial SHA-256 over the canonicalised core, packSha256 over the joined hashes, and every breakEvenTickets re-derived by ceiling division. Reports exactly which check failed. No Qloo call, no quota. Spends no Qloo calls.", "inputSchema": { "type": "object", "properties": { "packPath": { "type": "string" } }, "required": [], "additionalProperties": false } }, { "name": "offersheet_arm_gut", "description": "The Gut comparison arm: LLM alone, no external data. When no LLM provider is configured it returns status unavailable with no pick — never a rendered guess. Spends no Qloo calls.", "inputSchema": { "type": "object", "properties": { "question": { "type": "string" } }, "required": [ "question" ], "additionalProperties": false } }, { "name": "offersheet_arm_web", "description": "The Web comparison arm: LLM plus Wikidata facts only. When webFacts is absent and artist is given, facts are fetched via the Wikidata path first; empty facts still produce an honest run labeled with no Wikidata facts available, never invented facts. Spends no Qloo calls.", "inputSchema": { "type": "object", "properties": { "question": { "type": "string" }, "webFacts": { "type": "array", "items": { "type": "string" } }, "artist": { "type": "string" } }, "required": [ "question" ], "additionalProperties": false } }, { "name": "offersheet_miss_trial_build", "description": "Builds a miss-ledger trial from picks, purely offline. Returns null (no trial) unless a graph pick AND at least one comparison pick are present; identical picks yield a trial with empty misses. Spends no Qloo calls.", "inputSchema": { "type": "object", "properties": { "question": { "type": "string" }, "graphPick": { "type": "string" }, "gutPick": { "type": "string" }, "webPick": { "type": "string" } }, "required": [ "question", "graphPick" ], "additionalProperties": false } }, { "name": "offersheet_venue_search", "description": "Raw Ticketmaster venue lookup for a named room. A null result means the room did not resolve — never a guessed venue. This answers the lookup alone, without spending an events call. Spends no Qloo calls.", "inputSchema": { "type": "object", "properties": { "keyword": { "type": "string" } }, "required": [ "keyword" ], "additionalProperties": false } }, { "name": "offersheet_room_grounding", "description": "The composed Ticketmaster room leg: resolve a named room first, then scope the event lookup to its venue ID. Returns events found, venue name, capacity (value or unknown with its cause), price (value + source), and the grounding label — a resolved room's capacity is never borrowed from another room's event, and a first-event price/capacity is never silently labeled as the offer room's. Spends no Qloo calls.", "inputSchema": { "type": "object", "properties": { "artist": { "type": "string" }, "city": { "type": "string" }, "venue": { "type": "string" } }, "required": [ "artist", "city" ], "additionalProperties": false } }, { "name": "offersheet_deal_breakeven", "description": "Pure break-even arithmetic: tickets = ceil(guarantee / (ticketPrice - variableCost)); guarantee 0 gives 0 tickets with note no guarantee; net <= 0 is an error, never division by zero; tickets over a known capacity give Cannot break even in this room at this guarantee. A missing ticket price is an error, never an invented number. variableCostProvided keeps a blank cost distinct from zero cost. Spends no Qloo calls.", "inputSchema": { "type": "object", "properties": { "guarantee": { "type": "number", "minimum": 0 }, "ticketPrice": { "type": "number", "exclusiveMinimum": 0 }, "variableCost": { "type": "number", "minimum": 0 }, "capacity": { "type": "number", "exclusiveMinimum": 0 }, "variableCostProvided": { "type": "boolean" } }, "required": [ "guarantee", "ticketPrice" ], "additionalProperties": false } }, { "name": "offersheet_city_geocode", "description": "Keyless Nominatim city lookup (display name + lat/lon, or unavailable). Scope label on every response: display-only — shown for confirmation, never fed into Qloo filters. A failure never yields a guessed coordinate. Spends no Qloo calls.", "inputSchema": { "type": "object", "properties": { "query": { "type": "string" } }, "required": [ "query" ], "additionalProperties": false } }, { "name": "offersheet_artist_facts", "description": "Wikidata name-to-entity search only (id / label / description matches). Scope label on every response: input to the Web arm only, never a direct verdict input. No claim is ever derived from Wikidata alone. Spends no Qloo calls.", "inputSchema": { "type": "object", "properties": { "query": { "type": "string" }, "limit": { "type": "number", "minimum": 1 } }, "required": [ "query" ], "additionalProperties": false } }, { "name": "offersheet_health", "description": "Booleans only — qlooConfigured, ticketmasterConfigured, llmConfigured — plus status ok and version. No keys, no counts, no guard internals. Works with no keys configured; it is the tool that tells a client whether the others can work. Spends no Qloo calls.", "inputSchema": { "type": "object", "properties": {}, "required": [], "additionalProperties": false } }, { "name": "offersheet_limits_status", "description": "The server process's single Guards snapshot: monthlyBudget, monthlyCount, perVisitorDailyCap, perRunCallCap. These are OfferSheet's own adopted tripwires, NOT Qloo-stated quotas. Restart clears guard counters and the session receipt store. There is NO per-visitor usage lookup — Guards has no read method for it — so none is invented here. Spends no Qloo calls.", "inputSchema": { "type": "object", "properties": {}, "required": [], "additionalProperties": false } }, { "name": "offersheet_opener_shortlist", "description": "The composed opener chain: resolve the headliner, recommend opener candidates, then rank those candidates for the city audience — rank consumes recommend's output, so the steps stay sequential, never parallelised. Plus Gap-3 ID threading: an ID-less candidate gets one describe call before downstream use; a failed or ambiguous resolution keeps the name signal, never a guessed ID. Distinct from the plain verdict (no guarantee, no money, no full evidence legs, no venue/arms) and from calling recommend + rank by hand (the sequencing + ID threading is the logic). When recommend is empty, rank is honestly skipped, never an empty ranking presented as a pick. Spends Qloo quota (guarded).", "inputSchema": { "type": "object", "properties": { "artist": { "type": "string" }, "city": { "type": "string" }, "entityHint": { "type": "string" } }, "required": [ "artist", "city" ], "additionalProperties": false } }, { "name": "offersheet_opener_rerank", "description": "The HTTP POST /api/verdict/rerank endpoint as a tool, reusing its logic 1:1 (buildRerankArgs + one qloo_rank call + formatRerankResult, same receipt claim). Re-runs ONLY the ranking step with the same candidate set minus the excluded pick; the original verdict is never touched. An ok-but-empty rank envelope returns status 'empty' with the exact label and no pick — never an invented pick. excludedPick accepts a pick name string or an {id, name} object, as buildRerankArgs does. Spends Qloo quota (guarded).", "inputSchema": { "type": "object", "properties": { "artist": { "type": "string" }, "city": { "type": "string" }, "candidates": { "type": "array" }, "ranked": { "type": "array" }, "excludedPick": {}, "headlinerId": { "type": "string" } }, "required": [ "artist", "city", "excludedPick" ], "additionalProperties": false } } ] } } ], "toolCount": 33, "toolNames": [ "offersheet_verdict_run", "offersheet_verdict_race", "qloo_capabilities", "qloo_find_tags", "qloo_describe", "qloo_recommend", "qloo_rank", "qloo_where_popular", "qloo_audience_demographics", "qloo_entity_tags", "qloo_trends", "qloo_compare_audiences", "offersheet_rest_search", "offersheet_rest_insights_by_tags", "offersheet_rest_compare", "offersheet_rest_entities", "offersheet_receipt_replay", "offersheet_receipt_verify_hash", "offersheet_receipts_list", "offersheet_trials_list", "offersheet_trial_pack_verify", "offersheet_arm_gut", "offersheet_arm_web", "offersheet_miss_trial_build", "offersheet_venue_search", "offersheet_room_grounding", "offersheet_deal_breakeven", "offersheet_city_geocode", "offersheet_artist_facts", "offersheet_health", "offersheet_limits_status", "offersheet_opener_shortlist", "offersheet_opener_rerank" ], "offersheetTools": 23, "qlooTools": 10 }