{"server":"Lantern Works","checked":"2026-11-22","tools":[{"name":"compare_deals","description":"Compare 2-5 deals side by side (price, unit, offer type, code, verification), best value first.","inputSchema":{"type":"object","properties":{"ids":{"type":"string"},"mode":{"type":"string","default":"full"}},"required":["ids"],"additionalProperties":false},"annotations":{"readOnlyHint":true,"openWorldHint":true}},{"name":"compare_gpu_prices","description":"Compare tracked GPU rental prices for one model (e.g. 'H100', 'A100 SXM', 'RTX 4090', 'L40S'), cheapest first. If a result looks wrong, give the user its report_issue_url.\n\nSame data as https://thelanternworks.com/api/v1/gpu-prices.json. Sorted by hourly_usd ascending; each result\nhas source_url and checked_at. 'Cheapest' means cheapest among tracked providers, not the whole market.\nlimit 1-25. Unknown model returns not_found listing tracked models.\nH100/A100: cheapest_by_form_factor gives one pick per form_factor (tier_class, interruptible, in_stock); same rule as the site. 'NVLink' rows are form 'not stated': NVLink is the bridge, and the vendor doesn't say SXM or PCIe.\nEach result has tier_class: on-demand | community | marketplace | interruptible (spot/preemptible) |\nreserved | dedicated-monthly. Reserved (reservation/commitment 'starting from') rates are left out unless\ninclude_reserved=true, and even then they are never reported as cheapest. DEFAULT on_demand_only=true: results[0] is the cheapest\non-demand rate. Pass on_demand_only=false to include spot/interruptible, community and marketplace capacity (each\nsuch row is labelled by tier_class and interruptible); the summary then names the cheapest on-demand rate when the\noverall cheapest isn't. If a model has no on-demand rate, or every on-demand rate needs more GPUs than requested,\nthe default falls back to all rates and says so in on_demand_fallback.\ngpus (optional, 1-1024; default 1, stated in assumptions): the GPU count you need. A rate ranks only when its\nmin_gpus (smallest allocation sold) is <= gpus; rates sold only as larger allocations (e.g. 8-GPU nodes) are NOT\nranked and are listed in excluded_min_gpus (vendor, gpu, tier, min_gpus, gpu_count, hourly_usd, rate_id, source_url,\nchecked_at, reason, min_bill_usd_per_hour, min_bill_basis); min_gpus_unmet counts them (also in brief). Pass gpus=8 to\ncompare 8-GPU nodes. If every rate needs more GPUs, the error says which gpus value ranks them.\nEach result has min_bill_usd_per_hour = min_gpus x hourly_usd (min_bill_basis: hourly_rate, per_second_rate or\nmonthly_equivalent) and effective_at, when the vendor started charging this price (effective_at_basis vendor_stated,\nor first_observed = first daily snapshot showing it after a different price, an upper bound; null when not datable),\nseparate from last_verified. Neither affects ranking.\nFull results also include batch_due_ledger: each scheduled build (gpu-refresh, lium-collect, scout-import,\nscout-wider) has expected_at, last_expected_at, and the member_rate_ids it covers, derived only from the\nworkflow schedules. Brief mode omits the ledger.\nA failed or out-of-stock probe keeps failure_reason (sold_out, region, quota, or other) and\nfailure_reason_detail; both stay on a lapsed claim next to last_known_in_stock. Null when the source\ndoes not say; never guessed. Display only.\nA row with unknown stock, and each stale_prices entry, has next_check_at (next scheduled check, UTC), will_backfill\n(true: a scheduled check covers it; false: scope limit, nothing will re-check it; null: not derivable),\nbackfill_eta and next_check_basis (the schedule it came from).\ninclude_history=true adds each result's price_history (daily-snapshot change points) and history.recent_changes\n(price changes for this model among tracked providers, last 30 days) from /api/v1/gpu-prices/history.json.\nEach result has data_age_hours (hours since its price was last verified), degraded (> 48 h) and stale (> 7 days).\nA degraded row is shown and flagged but never ranked first while a fresh row exists; if none is fresh,\nstale_fallback=true and the summary says so.\nBy default a rate whose price is older than 168 h is never ranked or named cheapest: it is listed in stale_prices\n('stale price (N days old)'); if every rate is stale, results is empty and the summary says no current cheapest.\nmax_data_age_hours (optional) replaces the 168 h default and drops older rates; excluded_too_old counts them.\nregion (optional","inputSchema":{"type":"object","properties":{"gpu":{"type":"string"},"limit":{"type":"integer","default":10},"include_history":{"type":"boolean","default":false},"on_demand_only":{"type":"boolean","default":true},"include_reserved":{"type":"boolean","default":false},"max_data_age_hours":{"type":"number","default":null},"region":{"type":"string","default":null},"min_usable_memory_gb":{"type":"number","default":null},"gpus":{"type":"integer","default":null},"mode":{"type":"string","default":"brief"}},"required":["gpu"],"additionalProperties":false},"annotations":{"readOnlyHint":true,"openWorldHint":true}},{"name":"create_watch_key","description":"Hosted only, free, opt-in. Register a webhook for GPU price alerts and get an api_key plus a webhook signing secret (both shown ONCE; store them). webhook_url must be public https. email is optional contact only (no mail is sent). No payment. Stored: a hash of the key, the webhook URL, optional email, your watches and per-key counts.","inputSchema":{"type":"object","properties":{"webhook_url":{"type":"string"},"email":{"type":"string"}},"required":["webhook_url"]},"annotations":{"readOnlyHint":false,"openWorldHint":true}},{"name":"delete_watch","description":"Hosted only. Delete one of your watches by watch_id.","inputSchema":{"type":"object","properties":{"watch_id":{"type":"string"},"api_key":{"type":"string"}},"required":["watch_id"]},"annotations":{"readOnlyHint":false,"openWorldHint":true}},{"name":"estimate_job_cost","description":"Cost of one GPU job (not cost per hour) among tracked providers, cheapest current compute subtotal first.\n\ngpu_family (e.g. 'H100', 'A100 80GB') or its alias gpu (same name compare_gpu_prices uses) is required; if both\nare given they must match, else invalid_argument. gpus_per_job defaults to 1.\nRanking matches compare_gpu_prices' default: fresh rows first (a degraded price, older than 48 h, is shown and\nflagged but never ranked first while a fresh one exists), fresh on-demand before community/marketplace rows,\nthen compute subtotal ascending; the summary names any cheaper fresh rate ranked lower. If no fresh price exists,\nstale_fallback=true and the summary says the top row is not a current price.\n\nYou supply the duration: job_hours, OR work_units plus throughput_units_per_gpu_hour from your own benchmark\n(job_hours = work_units / (throughput x gpus_per_job)). If you give workload + work_units but no throughput,\na directly applicable published figure (closed-division MLPerf Inference, exact GPU variant, inference only)\nis used: throughput_basis='published_benchmark' with throughput_source {url, quote, date, benchmark, derived,\ncaveats}; your throughput always wins. Otherwise not_found 'missing: throughput for {gpu}'. Whole-instance billing:\ninstances = ceil(gpus_per_job / gpus_per_instance), all allocated GPUs billed (9 GPUs on an 8-GPU-only\noffer bill 16). Reserved and held-stale rates are never ranked; spot/interruptible only if allow_interruptible.\nEach result has rate_id, source_url, verified_at, data_age_hours/degraded/stale. Response has assumptions,\ncitations, missing_costs, budget_verdict (unresolved: extras unknown) and a quote_id. Compute subtotal only.","inputSchema":{"type":"object","properties":{"gpu_family":{"type":"string","default":null},"gpus_per_job":{"type":"integer","default":1},"job_hours":{"type":"number","default":null},"work_units":{"type":"number","default":null},"throughput_units_per_gpu_hour":{"type":"number","default":null},"workload":{"type":"string","default":null},"allow_interruptible":{"type":"boolean","default":false},"min_vram_gb":{"type":"integer","default":null},"runs_per_month":{"type":"number","default":null},"top_n":{"type":"integer","default":5},"max_data_age_hours":{"type":"number","default":null},"region":{"type":"string","default":null},"gpu":{"type":"string","default":null},"mode":{"type":"string","default":"full"}},"additionalProperties":false},"annotations":{"readOnlyHint":true,"openWorldHint":true}},{"name":"get_deal","description":"Get one deal's public record by id, including redeem_url and agent_instructions.","inputSchema":{"type":"object","properties":{"id":{"type":"string"},"mode":{"type":"string","default":"full"}},"required":["id"],"additionalProperties":false},"annotations":{"readOnlyHint":true,"openWorldHint":true}},{"name":"get_price_changes","description":"GPU rental price changes among tracked providers, each with old/new hourly price, change_pct, direction,\nthe vendor source_url, a stable event_id (dedupe key) and rate_id.\nMarketplace median moves also have composition_flag (composition, price, mixed, unknown, or null) from\nsuccessive median, minimum and listing count; composition, mixed and unknown are not price moves.\n\nsince: 'YYYY-MM-DD' returns every change on or after that date, newest first (daily snapshots in\n/api/v1/gpu-prices/history.json); omitted returns the latest day's moves (/api/v1/gpu-price-changes.json).\ncursor: pass next_cursor from a previous call to get every change after it, oldest first, with has_more;\nloop until has_more is false, and store next_cursor only after processing (a watcher never misses or repeats\nevents). error cursor_expired means re-sync with since and dedupe by event_id.\nkinds: optional subset of down, up, new, removed, same. gpu: optional model filter (substring, e.g. 'H100').\nlimit 1-200. Includes data_checked_at and next_refresh_utc (prices refresh daily at about 10:23 UTC).","inputSchema":{"type":"object","properties":{"since":{"type":"string","default":null},"gpu":{"type":"string","default":null},"limit":{"type":"integer","default":50},"cursor":{"type":"string","default":null},"kinds":{"type":"array","items":{"type":"string"},"default":null},"mode":{"type":"string","default":"full"}},"additionalProperties":false},"annotations":{"readOnlyHint":true,"openWorldHint":true}},{"name":"list_watches","description":"Hosted only. Your watches and per-key alert counts.","inputSchema":{"type":"object","properties":{"api_key":{"type":"string"}}},"annotations":{"readOnlyHint":true,"openWorldHint":true}},{"name":"optimize_agent_stack","description":"Estimate a monthly stack cost from a short description of the agent being built.\n\ndescription: free text, e.g. 'scrapes 20k JS pages/month via 5 GB residential proxies, 3k search queries,\nsummarizes with an LLM, budget $300/mo'. Parsed deterministically (keywords/regex, no LLM) into components:\ngpu, scraping, proxy, search, llm_api, email, hosting, database, monitoring. Optional fields override the parse:\ngpu_hours (GPU-hours/month), gpu_model ('H100', 'A100', 'RTX 4090'), model_size_b (billions of params; sets a\nminimum VRAM using FP16 x 1.2), scrape_pages, js_rendering, proxy_gb, search_queries, emails_sent, llm_api,\nmonthly_budget_usd. Each component gets the cheapest option with a cited price among tracked providers, with\ncalc, source_url, verified_at, referral_disclosure (set when the link is a referral link) and manual-signup\nnotes. Components with no cited price say 'no sourced price tracked' (nothing is estimated). Totals show the\nmath and say whether they are complete. Pulled listings are excluded. Every option has data_age_hours, degraded (> 48 h) and stale (> 7 days); GPU prices older than 168 h never rank by default (listed in the line's stale_prices); max_data_age_hours replaces that default and drops older options (excluded_too_old counts them). Same data as optimize_stack,\ncompare_gpu_prices and the deal feed.","inputSchema":{"type":"object","properties":{"description":{"type":"string","default":""},"workload":{"type":"string","default":null},"gpu_hours":{"type":"number","default":null},"gpu_model":{"type":"string","default":null},"model_size_b":{"type":"number","default":null},"scrape_pages":{"type":"number","default":null},"js_rendering":{"type":"boolean","default":null},"proxy_gb":{"type":"number","default":null},"search_queries":{"type":"number","default":null},"emails_sent":{"type":"number","default":null},"llm_api":{"type":"boolean","default":null},"monthly_budget_usd":{"type":"number","default":null},"exclude_manual_signup":{"type":"boolean","default":false},"max_data_age_hours":{"type":"number","default":null},"mode":{"type":"string","default":"full"}},"additionalProperties":false},"annotations":{"readOnlyHint":true,"openWorldHint":true}},{"name":"optimize_stack","description":"Cheapest suitable stack among tracked providers for monthly needs (set the quantities you need, rest 0).\n\nscrape_pages: pages/month without JS rendering. scrape_pages_js: pages/month needing JS rendering.\nproxy_gb: residential proxy GB/month. emails_sent: transactional emails/month. search_queries: search API\nqueries/month. gpu_hours: GPU-hours/month (1 GPU), optionally gpu_model e.g. 'H100', 'A100', 'RTX 4090'.\nEach line has the chosen plan, monthly_cost_usd, calc, vendor quote, source_url, checked_at and assumptions,\nplus alternatives. A vendor price that isn't published is 'unknown' (never estimated). manual_signup_only=true\nmeans a human must sign up (never automate it); exclude_manual_signup=true drops those plans. Same data as\nhttps://thelanternworks.com/api/v1/stack-pricing.json and /tools/stack-optimizer/.","inputSchema":{"type":"object","properties":{"scrape_pages":{"type":"integer","default":0},"scrape_pages_js":{"type":"integer","default":0},"proxy_gb":{"type":"number","default":0},"emails_sent":{"type":"integer","default":0},"search_queries":{"type":"integer","default":0},"gpu_hours":{"type":"number","default":0},"gpu_model":{"type":"string","default":null},"exclude_manual_signup":{"type":"boolean","default":false},"mode":{"type":"string","default":"full"}},"additionalProperties":false},"annotations":{"readOnlyHint":true,"openWorldHint":true}},{"name":"rebuild_quote","description":"Reopen a past quote. Pass the `receipt` object returned by optimize_stack, optimize_agent_stack or\nestimate_job_cost. Returns receipt_intact (unkeyed sha256 matches), reproduced (true only if each line's subtotal\nrecomputes AND its rate_id and price match the pinned feed revision or that date's published daily snapshot;\n\"arithmetic_only\" with provenance_verified=false when that can't be confirmed), per-line provenance, and\nrevalidation against today's feed: per GPU line unchanged | price_changed | evidence_changed | held_today |\nnot_in_feed_today (flat plans: not_revalidated), with revision 'same' or 'new'. Nothing is re-ranked.","inputSchema":{"type":"object","properties":{"receipt":{"type":"string"},"mode":{"type":"string","default":"full"}},"required":["receipt"],"additionalProperties":false},"annotations":{"readOnlyHint":true,"openWorldHint":true}},{"name":"search_deals","description":"Search agent work-stack deals (compute, search, email, unblock, hosting...).\n\noffer_type: verified_discount | listed_discount (not confirmed; verified_at null) | free_tier | trial | standard_referral | credit_only_referral.\nmax_price excludes deals with unknown price. limit 1-25. Pass next_cursor back as cursor to page.","inputSchema":{"type":"object","properties":{"query":{"type":"string","default":null},"category":{"type":"string","default":null},"offer_type":{"type":"string","default":null},"max_price":{"type":"number","default":null},"tags":{"type":"array","items":{"type":"string"},"default":null},"limit":{"type":"integer","default":10},"cursor":{"type":"string","default":null},"mode":{"type":"string","default":"full"}},"additionalProperties":false},"annotations":{"readOnlyHint":true,"openWorldHint":true}},{"name":"watch_price","description":"Hosted only. Alert your webhook (signed POST) after the daily refresh when a tracked rate for `gpu` crosses at or below max_hourly_usd, or a change of a watched kind (down/up/new; default down) lands at or below it. Only ELIGIBLE rates match (same contract as compare/estimate): every word of gpu must match the model; 'A100 80GB' or vram_gb=80 never matches a 40GB row; tiers (on-demand, community, marketplace, interruptible/spot, dedicated-monthly, reserved; default all but reserved) never matches another class; a row confirmed sold out never matches (in_stock='require' also skips unknown stock). region: code from availability.available_regions; a rate with no unexpired region evidence stays unresolved and never matches. Each event is delivered at most once per watch (dedupe on event_id). Pass api_key or an Authorization: Bearer header.","inputSchema":{"type":"object","properties":{"gpu":{"type":"string"},"max_hourly_usd":{"type":"number"},"region":{"type":"string"},"vram_gb":{"type":"integer"},"tiers":{"type":"array","items":{"type":"string"}},"in_stock":{"type":"string","enum":["exclude_sold_out","require"]},"kinds":{"type":"array","items":{"type":"string","enum":["down","up","new"]}},"api_key":{"type":"string"}},"required":["gpu","max_hourly_usd"]},"annotations":{"readOnlyHint":false,"openWorldHint":true}}]}