# Lantern Works

> Cited prices for AI agent infrastructure: GPU rental rates, scraping, proxies, email and search APIs, ranked by price for the buyer among tracked providers, not the whole market.

## Data
- [Agent discovery (JSON)](https://thelanternworks.com/agent.json): GPU-first entry points, hosted MCP, first call and decision rules; project-specific format
- [GPU model feed index (JSON)](https://thelanternworks.com/api/v1/gpu/index.json): choose a small model-specific feed instead of downloading every rate
- [GPU prices (JSON)](https://thelanternworks.com/api/v1/gpu-prices.json): per-row rate, tier, source_url, verified_at, availability
- [GPU coverage (JSON)](https://thelanternworks.com/api/v1/coverage.json): GPU model x vendor matrix (in stock, priced, not listed) with checked_at; page at https://thelanternworks.com/coverage/
- [Deals feed (JSON)](https://thelanternworks.com/api/v1/deals.json)
- [Stack pricing (JSON)](https://thelanternworks.com/api/v1/stack-pricing.json)
- [For agents](https://thelanternworks.com/for-agents/): endpoints, record shape, MCP server

## Availability rule
- A stock claim (availability.in_stock true or false) is valid until availability.availability_expires_at (UTC). Lambda, Verda, region-derived and per-region claims last at most 1 hour after checked_at; Vast marketplace overall claims last at most 1 hour; Clore overall claims at most 8 hours. After availability_expires_at, treat the claim as unknown: neither in stock nor sold out. Static feeds leave out claims within 25 minutes of expiry and are rebuilt on every availability snapshot. Consumers must still enforce availability_expires_at. compare_gpu_prices applies it at request time and is the source of truth.
- availability.in_stock null means no signal, not sold out.
- Pages show the same limit as "valid until HH:MM ET; after that treat as unknown".

## For agents
- [First cited quote, no installation](https://thelanternworks.com/quickstart/): Python 3.10+ standard-library MCP example; compare H100 SXM on-demand rates, then optionally estimate a job with a quote receipt. No API key for read-only queries.
- [Integration snippets](https://thelanternworks.com/integrate/): add remote MCP and a runnable first call to a README or dataset card, with a channel ref tag
- [Receipts MCP test summaries](https://thelanternworks.com/receipts/), [JSON](https://thelanternworks.com/api/v1/receipts.json): Pax-authored, dated task/version findings. Summary evidence only, not security certifications; deferred is untested, not broken. Separate from GPU job quote receipts.
- Prefer the MCP tool compare_gpu_prices in brief mode (mode='brief', the default): about 4k chars per answer with model, vendor, $/GPU-hr, checked_at and source_url. Hosted endpoint: https://lantern-works-mcp.lantern-works.workers.dev/mcp ; install: https://thelanternworks.com/install/
- Every page below embeds a machine-readable price block: <script type="application/json" id="lw-price-block">. GPU pages: items {model, vendor, usd_per_hr, quote_currency, checked_at, source_url}, sorted by cost, from gpu-prices.json (MCP: compare_gpu_prices). Non-GPU (tool/API) pages: items carry the deal's cited price_usd and unit with quote_currency where a price is cited; usd_per_hr is null with usd_per_hr_null_reason (these deals are not hourly); feeds deals.json and stack-pricing.json (MCP: compare_deals / search_deals); deals without a source_url and check date are left out of the block. Rows whose link carries a referral or affiliate code have referral_label "Referral link"; referral status never changes order.
- [Usage stats (JSON)](https://thelanternworks.com/stats.json): weekly aggregate ref= counts, no PII

## GPU price pages
- [A100 rental price](https://thelanternworks.com/gpu/a100-rental-price/)
- [B200 rental price](https://thelanternworks.com/gpu/b200-rental-price/)
- [B300 rental price](https://thelanternworks.com/gpu/b300-rental-price/)
- [H100 rental price](https://thelanternworks.com/gpu/h100-rental-price/)
- [H200 rental price](https://thelanternworks.com/gpu/h200-rental-price/)
- [L4 rental price](https://thelanternworks.com/gpu/l4-rental-price/)
- [L40S rental price](https://thelanternworks.com/gpu/l40s-rental-price/)
- [MI300X rental price](https://thelanternworks.com/gpu/mi300x-rental-price/)
- [RTX-4090 rental price](https://thelanternworks.com/gpu/rtx-4090-rental-price/)
- [RTX-5090 rental price](https://thelanternworks.com/gpu/rtx-5090-rental-price/)
- [RTX-6000-ADA rental price](https://thelanternworks.com/gpu/rtx-6000-ada-rental-price/)
- [RTX-A6000 rental price](https://thelanternworks.com/gpu/rtx-a6000-rental-price/)
- [V100 rental price](https://thelanternworks.com/gpu/v100-rental-price/)

## Reports
- [GPU price reports](https://thelanternworks.com/reports/): weekly, newest first
- [GPU Rental Price Report: Week of Nov 2, 2026](https://thelanternworks.com/reports/gpu-report-2026-11-02/)

## GPU comparison pages
- [All comparisons (index)](https://thelanternworks.com/compare/): GPU first; each page embeds lw-price-block JSON (price-block/1)
- [cheap-gpu-rental](https://thelanternworks.com/compare/cheap-gpu-rental/): GPU price block (model, vendor, usd_per_hr, checked_at, source_url)
- [cheapest-gpu-per-gb-vram](https://thelanternworks.com/compare/cheapest-gpu-per-gb-vram/): GPU price block (model, vendor, usd_per_hr, checked_at, source_url)
- [coreweave-vs-nebius-gpu-prices](https://thelanternworks.com/compare/coreweave-vs-nebius-gpu-prices/): GPU price block (model, vendor, usd_per_hr, checked_at, source_url)
- [lambda-vs-runpod-gpu-prices](https://thelanternworks.com/compare/lambda-vs-runpod-gpu-prices/): GPU price block (model, vendor, usd_per_hr, checked_at, source_url)
- [runpod-community-vs-secure-cloud-prices](https://thelanternworks.com/compare/runpod-community-vs-secure-cloud-prices/): GPU price block (model, vendor, usd_per_hr, checked_at, source_url)
- [runpod-vs-thunder-compute-vs-modal-h100](https://thelanternworks.com/compare/runpod-vs-thunder-compute-vs-modal-h100/): GPU price block (model, vendor, usd_per_hr, checked_at, source_url)
- [runpod-vs-verda-gpu-prices](https://thelanternworks.com/compare/runpod-vs-verda-gpu-prices/): GPU price block (model, vendor, usd_per_hr, checked_at, source_url)

## Tool and API comparison pages
- [ai-seo-writing-tools-comparison](https://thelanternworks.com/compare/ai-seo-writing-tools-comparison/)
- [best-rotating-proxies-for-web-scraping](https://thelanternworks.com/compare/best-rotating-proxies-for-web-scraping/)
- [cheapest-cloud-vps-for-ai-agents](https://thelanternworks.com/compare/cheapest-cloud-vps-for-ai-agents/)
- [cheapest-residential-proxies](https://thelanternworks.com/compare/cheapest-residential-proxies/)
- [cold-email-deliverability-warmup-tools](https://thelanternworks.com/compare/cold-email-deliverability-warmup-tools/)
- [email-verification-api-comparison](https://thelanternworks.com/compare/email-verification-api-comparison/)
- [screenshot-api-comparison](https://thelanternworks.com/compare/screenshot-api-comparison/)
- [serp-api-comparison](https://thelanternworks.com/compare/serp-api-comparison/)
- [transactional-email-api-pricing](https://thelanternworks.com/compare/transactional-email-api-pricing/)
- [web-scraping-api-deals](https://thelanternworks.com/compare/web-scraping-api-deals/)