# National Compute Docs

> Documentation for National Compute: burst GPU capacity rented through a market. You declare how many GPUs you want and the most you'll pay per GPU-hour; the market grants whole nodes and bills the clearing price, never above your ceiling.

Every page is public. Each link below is the page's markdown (its source, relative links made absolute); the HTML page is the same address without `.md` (`/market.md` -> `/market/`), and [llms-full.txt](https://docs.nationalcompute.com/llms-full.txt) is every page in one file. The sitemap is [sitemap.xml](https://docs.nationalcompute.com/sitemap.xml). The API itself is on https://nationalcompute.com: its orientation pages and OpenAPI contracts are listed under Machine surfaces, all without a token.

## Guides

- [Home](https://docs.nationalcompute.com/index.md): Burst GPU capacity allocated by price: declare the GPUs you want and the most you'll pay per GPU-hour; the market grants whole nodes at the clearing price.
- [Public Marshall](https://docs.nationalcompute.com/public-marshall.md): Model chat without sign-in, with a sponsored usage allowance per browser visitor; no shell, files, uploads or cluster access.
- [Sign in](https://docs.nationalcompute.com/sign-in.md): The one sign-in door (Google, an emailed sign-in link, or a password), who can create an account, public research organizations, and the waitlist.
- [Burst capacity](https://docs.nationalcompute.com/market.md): How burst capacity is allocated: the auction tick, the limit price, minimum-duration protection, base load capacity, and why scaling down is free.
- [Kubernetes clusters](https://docs.nationalcompute.com/kubernetes.md): A dedicated Kubernetes cluster whose GPU jobs create the demand: the limit price, launch admission, reclaim, LoadBalancer services, and the shared volume.
- [Slurm clusters](https://docs.nationalcompute.com/slurm.md): A dedicated Slurm cluster: the login node, jobs and hooks, containers, storage, controller failover, how compute nodes come and go, and monitoring.
- [Public Research](https://docs.nationalcompute.com/public-research.md): One whole GPU node at a time for a researcher at a public research organization: the queue, the lease, working over ssh, and billing at a flat rate.
- [Billing](https://docs.nationalcompute.com/billing.md): What the meter measures and what is never billed: prepaid credits, the clearing price for capacity held, base load, Marshall model usage, invoices.
- [Base Load Capacity SLA](https://docs.nationalcompute.com/sla.md): The base load capacity uptime commitment per GPU node, how uptime is measured, and the credit schedule.
- [Authentication](https://docs.nationalcompute.com/authentication.md): Org API tokens for the REST APIs and OAuth sign-in for the MCP server: minting, token-to-cluster resolution, discovery, auth errors, and what needs no auth.
- [Trust Center](https://docs.nationalcompute.com/trust.md): Tenant isolation by whole nodes in dedicated clusters, node sanitization, audit records, API credentials, data sharing, and what the platform does not claim.
- [Marshall data sharing](https://docs.nationalcompute.com/marshall-data-sharing.md): The organization-level data sharing tiers for signed-in Marshall sessions: what each tier collects, changing it, revocation, retention, and redaction.
- [FAQ](https://docs.nationalcompute.com/faq.md): Answers on allocation and base load, paying for what you use, the pricing index, auction mechanics, the access model, and storage costs.

## API reference

- [Conventions](https://docs.nationalcompute.com/api/index.md): API conventions on the apex host: the OpenAPI contracts, the error shape, versions and compare-and-set, rate limits, prices, nodes and GPUs.
- [MCP server](https://docs.nationalcompute.com/api/mcp.md): The hosted MCP server: connecting with OAuth sign-in, the tools, grid_api paths, two-phase confirmation of spending tools, feedback rewards, and limits.
- [Telemetry and flags](https://docs.nationalcompute.com/api/metrics.md): metrics_read: per GPU telemetry and the platform's flags on it for the hardware a job ran on; metrics_query for free-form PromQL.
- [Slurm reads](https://docs.nationalcompute.com/api/slurm.md): slurm_read: a Slurm cluster's status page, one job, its place in the queue and its log, read only through the MCP server.
- [Dashboards](https://docs.nationalcompute.com/api/dashboards.md): dashboard_write turns PromQL into a Grafana dashboard document in your organization's Grafana; dashboard_read lists them.
- [Checkpoints](https://docs.nationalcompute.com/api/checkpoints.md): Checkpoints: static files published by an organization's members, kept and served back; the public feed, the authenticated routes, the MCP tools.
- [Hardware roadmap](https://docs.nationalcompute.com/api/roadmap.md): roadmap_read: the calendar quarter in which a GPU model is expected on the platform, an expectation and never a commitment.
- [VM capacity](https://docs.nationalcompute.com/api/vm-capacity.md): The VM capacity API: register an ssh key, declare GPUs and a price ceiling, read state, the allocation lifecycle, swapping a node, history, and errors.
- [Kubernetes limit price](https://docs.nationalcompute.com/api/k8s-bid.md): The Kubernetes limit price API: one number per cluster or per job, how the bid prices the market, queue events, shared storage volumes, and tokens.
- [Market analytics](https://docs.nationalcompute.com/api/k8s-market.md): Five read only routes for a Kubernetes cluster's market: the price to win ladder, what held nodes are assessed, your limit price over time, demand by band.
- [Cluster facts and Base Load](https://docs.nationalcompute.com/api/k8s-cluster.md): Three read only routes: a Kubernetes cluster's facts, its kubeconfig, and the base load book it can hold.
- [Cluster pages by token](https://docs.nationalcompute.com/api/k8s-cluster-pages.md): Thirteen read only routes serving the console's Workloads, job, Cluster Overview, machine and Storage pages for a Kubernetes cluster, by org API token.
- [Price feed](https://docs.nationalcompute.com/api/market-feed.md): The price feed: per-tick clearing prices and bucketed history, read only, and how to tune a ceiling against the going rate.
- [Billing records](https://docs.nationalcompute.com/api/billing.md): Billing records by org API token: the ledger per node and price window, daily totals, balance and burn rate, spend per workload, prices in force, writes.
- [Organization](https://docs.nationalcompute.com/api/organization.md): The organization routes: members and invitations, API tokens as metadata, each token's activity trail, and revoking your own token.
- [Errors](https://docs.nationalcompute.com/api/errors.md): Every error code the API returns, grouped: auth and addressing, validation, concurrency and state, pacing and platform, admission denials, handling.

## Machine surfaces

- [llms.txt](https://nationalcompute.com/llms.txt): the API's own orientation: the two capacity surfaces, pricing, the refusals that matter
- [AGENTS.md](https://nationalcompute.com/AGENTS.md): orientation for an agent driving the API or the MCP server
- [FIRST-JOB.md](https://nationalcompute.com/FIRST-JOB.md): the first job, agent shaped: connect, price, launch and watch
- [VM capacity OpenAPI](https://nationalcompute.com/api/vm/openapi.json): the VM capacity contract
- [Kubernetes limit price OpenAPI](https://nationalcompute.com/api/k8s/openapi.json): the limit price contract for Kubernetes clusters
- [Market feed OpenAPI](https://nationalcompute.com/api/market/openapi.json): the clearing-price feed contract
- [Billing OpenAPI](https://nationalcompute.com/api/billing/openapi.json): the ledger, balance and spend-limit contract
- [Organization OpenAPI](https://nationalcompute.com/api/org/openapi.json): the roster, invitations and API token contract
- [MCP server](https://nationalcompute.com/mcp): the tenant surface as MCP tools over Streamable HTTP, OAuth sign-in