llms.txt
+36 −3
https://nodehour.forevenza.com/llms.txt
− > Nodehour helps teams source GPU and accelerator capacity: chip type, count, hours or term, region, interconnect and how firmly they would commit, specified so a provider can say at once whether it has the configuration. Nodehour looks at every requirement and replies when it can find a provider with that configuration or get it quoted; nothing is reserved until the buyer agrees terms.− ## Services− - [GPU and accelerator capacity by type, count and term](https://nodehour.forevenza.com/compute-capacity/index.md): Source GPU or accelerator capacity: chip type, count, hours or term, region, interconnect and how firmly you would commit, specified so a provider can answer at once.+ > A buyer can describe a training run, fine-tune or serving target in plain words (model size, tokens or tokens/s, start, deadline, region, stack) or give a known cluster spec. The page proposes a configuration with its GPU-hours arithmetic, one alternative and the decisions that change the quote, then sends the requirement. Each request gets a reference at once; Nodehour takes the configuration to providers, asks each to quote the same terms and replies by email with options side by side. Nothi+ ## Pages+ - [Nodehour: describe the run, get the cluster sized and quoted](https://nodehour.forevenza.com/index.md): Describe a training run or serving fleet in plain words. Nodehour sizes the GPU cluster with arithmetic you can check and brings back provider quotes side by side.+ - [GPU capacity, sourced against your requirement | Nodehour](https://nodehour.forevenza.com/compute-capacity/index.md): What Nodehour sources (chips, fabrics, delivery, terms) and how a requirement becomes options, a choice and an order placed with the provider.+ - [Sizing method: GPU-hours, nodes and serving memory | Nodehour](https://nodehour.forevenza.com/sizing/index.md): Every formula behind Nodehour's sizing sheet, the chip figures with vendor sources, and what the estimate does not cover.+ - [For agents: submit a GPU capacity requirement | Nodehour](https://nodehour.forevenza.com/for-agents/index.md): The request envelope, a minimal example with partial specifications, and what the response contains.+ ## What you can do+ - Send a requirement: POST https://nodehour.forevenza.com/_c/request with the envelope in https://nodehour.forevenza.com/openapi.json (or the MCP tool submit_request). You get a reference number at once, and we keep exactly what you sent: your words, the proposed configuration, the terms to quote and the open questions. If sizing or an open question changes the quote we email you first; then we take the configuration to providers that can run it, ask each to quote the same terms, and send the op+ - Resume a requirement with request_status using its receipt id and token. Where the business answers a request that left no address, it answers on the request itself (not every request can be answered that way, and an address, where the buyer has one, is the surer way to hear back). Its replies (messages) and any linked authorized offer appear in request_status, so read it again later. Answer a reply, or send additional requirements, with continuation: {id, token}. A figure in a reply is indica+ ## Connect from your agent+ The MCP endpoint is https://nodehour.forevenza.com/mcp (Streamable HTTP, stateless, no sign-in). Adding it in one host does not add it in another.+ - Claude (claude.ai, desktop and mobile): In Claude, open Customize, then Connectors, choose Add custom connector and enter https://nodehour.forevenza.com/mcp with no sign-in.+ - ChatGPT: In ChatGPT on the web, open Plugins, choose + and Add custom MCP server, and enter https://nodehour.forevenza.com/mcp with no authentication, where your plan allows custom servers.+ - Claude Code: Add it from a terminal:+ ```+ claude mcp add --transport http nodehour https://nodehour.forevenza.com/mcp+ ```+ - Codex: Add it to ~/.codex/config.toml:+ ```+ [mcp_servers.nodehour]+ url = "https://nodehour.forevenza.com/mcp"+ ```+ - Gemini CLI: Add it from a terminal:+ ```+ gemini mcp add --transport http nodehour https://nodehour.forevenza.com/mcp+ ```+ - Antigravity: Add it from a terminal:+ ```+ agy mcp add nodehour https://nodehour.forevenza.com/mcp+ ```+ - Grok (grok.com and Grok CLI): On grok.com add a New Connector, Custom, with https://nodehour.forevenza.com/mcp; in Grok CLI:+ ```+ grok mcp add --transport http nodehour https://nodehour.forevenza.com/mcp+ ```+ - VS Code (GitHub Copilot): Open the install link in VS Code, or run MCP: Add Server and enter https://nodehour.forevenza.com/mcp. vscode:mcp/install?%7B%22name%22%3A%22nodehour%22%2C%22type%22%3A%22http%22%2C%22url%22%3A%22https%3A%2F%2Fnodehour.forevenza.com%2Fmcp%22%7D+ - Cursor: Open the install link in Cursor, or add https://nodehour.forevenza.com/mcp under MCP settings. cursor://anysphere.cursor-deeplink/mcp/install?name=nodehour&config=eyJ1cmwiOiJodHRwczovL25vZGVob3VyLmZvcmV2ZW56YS5jb20vbWNwIn0%3D