Lantern Works agent deals
# Lantern Works
Lantern Works compares live GPU rental prices across tracked providers, for agents and people, through a free MCP server and the site.

```bash
uvx lantern-works-deals-mcp
```
[thelanternworks.com](https://thelanternworks.com) · [Receipts](https://lantern-works-mcp.lantern-works.workers.dev/receipts/?ref=receipts-readme)
## Get your first quote — no installation or API key
Python 3.10+, standard library only. [Read the example](examples/first_quote.py), then download and run it:
```sh
curl --fail --location https://raw.githubusercontent.com/connormatthewdouglas/agent-work-stack-deals/main/examples/first_quote.py --output first_quote.py
python3 first_quote.py --gpu 'H100 SXM' --hours 100 --ref readme
```
This calls the public hosted MCP: three on-demand rates checked within 24 hours, then a 100-hour, one-GPU compute estimate with sources, assumptions, missing costs and a receipt. It creates no purchase or watch. Use `--qa` for internal tests.
[Client setup](https://thelanternworks.com/install/?ref=readme) · [Quickstart](https://thelanternworks.com/quickstart/?ref=readme) · [Agent manifest](https://thelanternworks.com/agent.json) · [Small per-model JSON feeds](https://thelanternworks.com/api/v1/gpu/index.json) · [Add this to your README or dataset](docs/INTEGRATING.md)
- **GPU prices:** [thelanternworks.com](https://thelanternworks.com/?ref=readme) · [/gpu/](https://thelanternworks.com/gpu/?ref=readme) · JSON [api/v1/gpu-prices.json](https://thelanternworks.com/api/v1/gpu-prices.json) · history [api/v1/gpu-prices/history.json](https://thelanternworks.com/api/v1/gpu-prices/history.json) · what moved since yesterday [api/v1/gpu-price-changes.json](https://thelanternworks.com/api/v1/gpu-price-changes.json) ([RSS](https://thelanternworks.com/api/v1/gpu-price-changes.xml))
- **GPU coverage:** which GPU models each vendor is tracked for (in stock, priced, not listed): [/coverage/](https://thelanternworks.com/coverage/?ref=readme) · JSON [api/v1/coverage.json](https://thelanternworks.com/api/v1/coverage.json)
- **Deals directory:** [thelanternworks.com/deals/](https://thelanternworks.com/deals/?ref=readme) · [About](https://thelanternworks.com/lanternworks/?ref=readme)
- **Agent feed (JSON, no auth):** [thelanternworks.com/api/v1/deals.json](https://thelanternworks.com/api/v1/deals.json) (same bytes as `/deals.json`) · mirror: [gist](https://gist.githubusercontent.com/connormatthewdouglas/ab5c3a8dceaf396d8f0c631913433a5e/raw/deals.json)
- **Stack pricing:** [api/v1/stack-pricing.json](https://thelanternworks.com/api/v1/stack-pricing.json)
- **Query CLI:** `python3 query_deals.py --category unblock --max-price 20` (filter contract: [USAGE.md](USAGE.md))
- **MCP test summaries:** [Receipts](https://lantern-works-mcp.lantern-works.workers.dev/receipts/?ref=receipts-readme) · [JSON](https://lantern-works-mcp.lantern-works.workers.dev/api/v1/receipts.json?ref=receipts-readme) (dated findings, separate from job-price receipts)
Contact: [lanternworks@agentmail.to](mailto:lanternworks@agentmail.to)
## MCP server (v0.6.3, read-only, stdio)
Free, read-only, no API key. Package [`lantern-works-deals-mcp`](https://pypi.org/project/lantern-works-deals-mcp/) on PyPI; MCP Registry name `io.github.connormatthewdouglas/lantern-works-deals`. Built on the official Python MCP SDK.
Tool counts: **9** in the local package (stdio, below); **13** on the hosted endpoint (the same 9 plus `create_watch_key`, `watch_price`, `list_watches`, `delete_watch`).
| Tool | What it does |
|---|---|
| `compare_gpu_prices` | Hourly rental prices for a GPU model across tracked providers, cheapest first. When a result includes `availability_rate_7d`, `window_days` is the planned window and `observed_hours` is the actual coverage (`coverage_note` when coverage is shorter). |
| `get_price_changes` | GPU price moves among tracked providers, newest first. |
| `optimize_stack` | Cheapest tracked plan per usage line, with sources. |
| `optimize_agent_stack` | Cheapest cited option per component from a plain-text agent description. |
| `estimate_job_cost` | Cost of one GPU job among tracked providers, cheapest first. |
| `rebuild_quote` | Reopen a past quote and revalidate it against today's feed. |
| `search_deals` | Search the deals feed by text, category, offer type, price and tags. |
| `compare_deals` | Compare 2-5 deals by id, best customer value first. |
| `get_deal` | Return the full public record for one deal id. |
The full tool contracts, eligibility and stale-price rules, hosted endpoint, alert delivery, throughput data, and the feed contract and deploy gate are in [docs/CONTRACT.md](docs/CONTRACT.md). Arguments, errors, data sources and tests: [mcp_server/README.md](mcp_server/README.md).
### Install
```bash
uvx lantern-works-deals-mcp
```
Quickstart: ask for `compare_gpu_prices(gpu="H100")`, then the next day call `get_price_changes(since="<data_checked_at date>")` to see what moved. Prices refresh daily at about 10:23 UTC; `compare_gpu_prices` and `optimize_agent_stack` return `data_checked_at` and `next_refresh_utc`.
Claude Code:
```bash
claude mcp add lantern-works-deals -- uvx lantern-works-deals-mcp
```
Cursor (`.cursor/mcp.json`) and Cline (`cline_mcp_settings.json`, via MCP Servers → Configure) use the same entry:
```json
{ "mcpServers": { "lantern-works-deals": { "command": "uvx", "args": ["lantern-works-deals-mcp"] } } }
```
Docker: `docker build -t lantern-works-deals-mcp . && docker run -i --rm lantern-works-deals-mcp` (the root [`Dockerfile`](Dockerfile) installs the pinned PyPI release and runs the stdio server). Without uv: `pip install lantern-works-deals-mcp`, then use `"command": "lantern-works-deals-mcp"`. Other clients: https://thelanternworks.com/install/?ref=readme · Skill: `npx skills add connormatthewdouglas/agent-work-stack-deals`.
Telemetry is **off by default**. `LANTERN_TELEMETRY=1` opts in to an anonymous daily count of tool calls: tool name, version, day and count only. No arguments, IDs or IPs are sent or stored. See [PRIVACY.md](PRIVACY.md).
Hosted endpoint (no install): `https://lantern-works-mcp.lantern-works.workers.dev/mcp?ref=readme`. Privacy and caller-count detail: [docs/CONTRACT.md](docs/CONTRACT.md), [PRIVACY.md](PRIVACY.md) and [/privacy/](https://thelanternworks.com/privacy/).

Demo (live data, 2026-10-28): the MCP server answers "cheapest H100 for 100 hours?". It shows 100-hour totals for the cheapest, next and most expensive tracked rates, each with its source and check time, and labels referral links. Recording: [docs/demo/cheapest-h100-100h.cast](docs/demo/cheapest-h100-100h.cast) · script: [docs/demo/demo.py](docs/demo/demo.py).
Framework examples: [LangGraph](examples/langgraph-cheapest-h100/) (cheapest H100), [CrewAI](examples/crewai-cheapest-gpu/) and [OpenAI Agents SDK](examples/openai-agents-cheapest-gpu/) (cheapest GPU for a described job), [Stack optimizer](examples/stack-optimizer/) (describe your agent, get the cheapest cited stack via `optimize_agent_stack`), [GPU price watcher](examples/gpu-price-watcher/) (what moved since a date via `get_price_changes`, cron-ready), [GPU price monitor](examples/gpu-price-monitor/) (price-drop alerts from the history API). All examples: [examples/](examples/). Some tool results include referral links (`affiliate: true`).
## Report a wrong price
See a price that doesn't match the vendor's page, or an offer you can't rent? [Open a wrong-price report](https://github.com/connormatthewdouglas/agent-work-stack-deals/issues/new?template=wrong-price.yml). Agents can prefill it from a feed row; see [docs/REPORT-WRONG-PRICE.md](docs/REPORT-WRONG-PRICE.md).
## License
- Code: MIT, see [LICENSE](LICENSE).
- Field reference for `canonical_model`, `form_factor`, and `availability_rate_7d` (`window_days` = planned window, `observed_hours` = actual coverage) in `gpu-prices.json`: [docs/GPU-PRICES-FIELDS.md](docs/GPU-PRICES-FIELDS.md).
- Price and deal data (`deals.json`, `api/v1/*.json` including `gpu-prices.json`, the price history and change feed, `data/gpu_pricing/`): [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/), see [LICENSE-DATA](LICENSE-DATA). Attribution: "Lantern Works" with a link to https://thelanternworks.com. Every JSON feed carries these terms in its `license` field.
TDQS
Scored across 7 tools
Most tools have distinct resource+action targets: search_deals finds deals, compare_deals compares specific ones, get_deal fetches a single record, and compare_gpu_prices vs get_price_changes separate snapshot pricing from movement history. The one real overlap is optimize_stack vs optimize_agent_stack, since the latter accepts all of the former's explicit fields as overrides, making the plain optimizer largely a subset.
Every tool follows a clean snake_case verb_noun pattern (compare_*, get_*, search_*, optimize_*), with verbs that accurately signal the operation. No mixed casing or inconsistent verb styles anywhere in the set.
Seven tools is well-scoped for a price/deal data service, with each covering a distinct facet: GPU price comparison, price movement, deal search, deal comparison, deal lookup, and two stack cost estimators. Nothing feels padded or missing at the count level.
The surface covers the core workflows: find/compare/get deals, compare GPU prices, track changes, and estimate stack costs with history via parameters. Minor gaps remain — no way to list tracked models or providers directly (only via not_found or a substring filter), and no per-provider price table across models, but agents can work around these.