Add cited GPU prices to your agent or README

Give builders a useful next step after discovering your GPU tool, dataset or agent guide: compare rental costs with source links, then estimate their job. Free read-only queries, no signup.

Start with one direct price link. Replace partner-readme with a short channel tag (lowercase letters, digits, underscores or hyphens; at most 32 characters). The tag records aggregate channel counts; it does not identify a user or prove an installation.

Lantern Works tracks [on-demand GPU rental prices](https://thelanternworks.com/gpu/price-comparison/?ref=partner-readme) across multiple providers, with vendor sources and check times.

Optional runnable MCP integration (Markdown)

### Compare GPU rental costs in your agent

[Lantern Works](https://thelanternworks.com/quickstart/?ref=partner-readme) compares GPU rental prices among tracked
providers, with vendor sources and check times. Free read-only queries; no API key.
Useful before a training run or when checking a job's compute budget.

For MCP clients that accept a remote `mcpServers` URL (see [client-specific setup](https://thelanternworks.com/install/?ref=partner-readme)):

```json
{
  "mcpServers": {
    "lantern-works": {
      "url": "https://lantern-works-mcp.lantern-works.workers.dev/mcp?ref=partner-readme"
    }
  }
}
```

Ask: `compare_gpu_prices(gpu="H100 SXM", on_demand_only=true, limit=3, mode="brief")`.
For a job: `estimate_job_cost(gpu_family="H100 SXM", gpus_per_job=1, job_hours=100)`.

Try it without an MCP client (Python 3.10+, standard library only; download and inspect the example first):

```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 partner-readme
```

Use `source_url` and `checked_at` from each result. Cheapest means among tracked
providers, not the entire market. Stock can be unknown; estimates are compute
subtotals and list missing costs. Referral links are labeled and never affect ranking.
[Small per-model JSON feeds](https://thelanternworks.com/api/v1/gpu/index.json) are available without MCP.

Generate a placement-specific version

In the source repository, run python3 scripts/integration_snippet.py --ref your-project. Add --format quickstart for the runnable example; use --ref hf or --ref kaggle for dataset cards.

Keep the source/freshness and missing-cost limits beside the example. Use the JSON model index when your application does not support MCP. Client configuration formats vary: link to the client-specific install guide.

Try the first quote · Check model coverage · Price methodology