Tensorust

Datos centrados en recibos para agentes: descargas reales de crates.io (bots del día de lanzamiento eliminados), precios de API de LLM con fecha y el modelo más barato para una tarea, respuestas con fuentes y 14 herramientas de desarrollo gratuitas (cron, semver, UUID, JWT, tokens, VRAM). Gratis, sin clave.

Servidor MCP alojado

npx add-mcp 'https://mcp.tensorust.com/mcp'

Se instala en Claude Code, Codex, Cursor y más

Documentación

Developers · free tier · keys for history

Use Tensorust from your code or your agent.

Everything on this site, as data and tools: real Rust crate downloads, LLM API prices, The Manual's answers and our free calculators. One REST API, one MCP server, plain JSON files. Current values are free with no key; the full history behind them takes a key or a few cents per call.

Short answer: agents connect to https://mcp.tensorust.com/mcp (MCP, Streamable HTTP). Code calls https://api.tensorust.com/v1. Bulk data is static JSON at https://tensorust.com/data/v1/. Current values are free at 60 requests a minute per IP (MCP: 60 tool calls); the static files have no limit. Full history (crate downloads, AI prices, GPU prices, uptime) with CSV is $19 a month (get a key), or $0.01 a call over HTTP 402 for agents.

https://mcp.tensorust.com/mcp MCP server for Claude, Cursor, VS Code, any agent

https://api.tensorust.com/v1 REST API, JSON, CORS open

/data/v1/ 1000 crates, 71 models, 302 pages, 24 tools

Try it now

Search everything we publish. This calls the live API from your browser.

Connect an agent (MCP)

Add the server once and your agent can look up crates, price models, read The Manual and run our tools in the middle of a task.

claude mcp add --transport http tensorust https://mcp.tensorust.com/mcp
ToolWhat it doesExample arguments
search_siteSearch everything we publish: crates, models, Manual pages, articles, reports, tools.{"query": "uuidv7"}
get_crateOne Rust crate: real downloads a day, release-window share, trend, releases.{"name": "serde"}
compare_cratesTwo to ten crates side by side, with a plain verdict.{"names": ["serde", "tokio"]}
get_model_priceOne LLM API model: every price tier, workload costs, history. Optional token counts.{"model": "claude-sonnet-5-5", "input_tokens": 2000, "output_tokens": 500}
pick_model_for_jobCheapest models for a job (chat, rag, sum, agent, batch, long) or your own token counts.{"job": "rag", "requests_per_month": 50000}
get_manual_pageA Manual page or article: short answer, key facts with sources, FAQ.{"slug": "uuidv7-vs-uuidv4"}
money_sizeAny dollar amount as cash: bills, weight, volume, pallets, what it fills.{"amount": 1000000000}
run_toolRun a free tool: llm_cost, count_tokens, cron_explain, semver_check, uuid_decode, snowflake_decode, unix_timestamp, jwt_decode, json_repair, llm_vram, app_store_fee, steam_revenue.{"tool": "semver_check", "args": {"requirement": "^1.2", "versions": ["1.1.9", "1.4.0", "2.0.0"]}}
crate_checkIs a Rust crate safe, is my version affected, and the exact cargo command to fix it (live RustSec/OSV + crates.io).{"name": "time", "version": "0.2.22"}
gpu_cheapestCheapest rental for one GPU right now, monthly cost, spot option, where to rent.{"gpu": "H100"}
compare_models2 to 10 LLM APIs priced on one workload, with a verdict and what to switch to.{"models": ["gpt-5-mini", "gemini-2.5-flash"], "job": "rag"}
search_articles / list_datasetsArticles with summary, URL and data file URLs; every article CSV with columns and rows.{"q": "gpu", "with_data": true}
list_reportsReport series and issues with short answers and CSV links.{}
list_toolsEvery free tool on the site, and which ones run_tool runs.{}
gpu_prices / service_statusFree: GPU rental prices today, and whether 14 developer services are up, with 24h/7d/30d uptime.{"gpu": "H100"}
history_crate_downloadsKey: every day of crate downloads we have kept, past the 90 days crates.io shows.{"api_key": "tdk_...", "name": "tokio"}
history_llm_prices / history_gpu_prices / history_uptimeKey: full AI price, GPU price and uptime history.{"api_key": "tdk_...", "model": "gpt-5-mini"}
floor_posts / builds_listThe Floor and Agent Builds. Not live yet: they say so until the community API ships.{}

Call the REST API

Cheapest model for a chatbot at 100,000 replies a month

curl "https://api.tensorust.com/v1/llm/cheapest?job=chat&requests_per_month=100000&limit=3"

Which crate do people really use? Compare real downloads

curl "https://api.tensorust.com/v1/crates/compare?names=serde,tokio,anyhow"

Run a tool: cron in plain English with the next runs

curl -X POST "https://api.tensorust.com/v1/tools/cron_explain" \
  -H "content-type: application/json" \
  -d '{"expression": "*/15 9-17 * * 1-5", "timezone": "America/New_York"}'
MethodPath (base https://api.tensorust.com/v1)What
GET/search?q=Search everything
GET/cratesList crates (sort, category, q, limit)
GET/crates/{name}One crate with daily numbers
GET/crates/compare?names=a,bCompare 2 to 10 crates
GET/llm/modelsList models (provider, q)
GET/llm/models/{id}One model, optional cost for your tokens
GET/llm/cheapest?job=Cheapest models for a job or token counts
GET/llm/changesDated price changes
GET/manualManual index (q, kind, topic)
GET/manual/{slug}One Manual page
GET/articlesArticle index
GET/articles/{slug}One article
GET/reportsReports and issues
GET/crates/{name}/check?version=Crate verdict + advisories; is your version affected, fix version, cargo command
GET/advisories/{name}/{version}Same check, version in the path
GET/gpu/{gpu}/cheapestCheapest rental for one GPU now, where to rent (free)
GET/llm/compare?models=a,bPrice 2 to 10 models on one workload
GET/articles/search?q=Articles with summary, URL, data file URLs
GET/datasetsEvery article data file (CSV), columns, rows
GET/money?amount=Money size for any amount
GET/toolsFree tools, and which can run here
POST/tools/{tool}Run a tool with a JSON body (GET with query params works too)
GET/gpuGPU rental prices today (free)
GET/statusDeveloper services up or down now, with uptime (free)
GET/historyHistory datasets and prices (free)
GET/history/crates/{name}Key or $0.01: every day of downloads we have kept
GET/history/llm[/{model}]Key or $0.01: AI API price history
GET/history/gpu[/{gpu}]Key or $0.01: GPU price history
GET/history/uptime[/{service}]Key or $0.01: uptime by day, hour or 5-minute check
GET/bulk/cratesBusiness key or $0.10: every crate in one file
GET/openapi.jsonOpenAPI 3.1 spec

Full spec: OpenAPI 3.1 (also at /developers/openapi.json). Import it into Postman, Insomnia or a GPT action.

Pricing

PlanPriceWhat you get
Free$0, no keyCurrent values for every dataset, 90 days of daily crate downloads, every tool. 60 requests a minute per IP.
Pay per call$0.01 a call ($0.10 bulk)Any history endpoint with no account: it answers HTTP 402 with x402 terms (USDC on Base); sign, retry, get the data. Nothing is charged for a failed call.
Pro$19 a monthFull history for every crate, model, GPU and service, CSV export, 600 requests a minute, the same key on MCP.
Business$49 a monthPro plus bulk endpoints (every crate in one file), 3,000 requests a minute, and commercial use of the data in your product or client work.

Get a Pro or Business key →

The key appears right after paying; cancel anytime from the same page.

Call a history endpoint with your key

# Pro or Business key from https://shop.tensorust.com/dev/data-api
curl -H "Authorization: Bearer $TENSORUST_KEY" \
  "https://api.tensorust.com/v1/history/crates/serde?format=csv" -o serde.csv

Send the key as Authorization: Bearer tdk_... or X-API-Key. Add ?format=csv for a spreadsheet, and from / to (YYYY-MM-DD) to cut the range. Crate downloads go back to Jul 27, 2026 and grow every day, past the 90 days crates.io itself keeps; AI prices start Oct 17, GPU prices and uptime Oct 24, 2026.

Bulk data (static JSON)

Every file is rebuilt with the site and cached on the CDN. Use these when you want everything at once; there is no rate limit. Start at /data/v1/index.json.

What's free, and limits

  • Free: current values, search, The Manual, reports and every tool. No key, no account.
  • Paid: only the full history (/history/*, /bulk/*). See pricing.
  • Rate limit: 60 requests a minute per IP on the API, and 60 tool calls a minute on the MCP server (connecting and listing tools are not counted). Over that you get HTTP 429 with a Retry-After header. Need more? Use the static JSON, or email matt@tensorust.com.
  • Agents pay per call through HTTP 402 (x402, USDC on Base) with no account, settled only when the data is ready. Anything free now stays free.
  • Credit: data is CC BY 4.0. Link the page the number came from (every record has a url).
  • Receipts: every price and number carries its source and the date we checked it. If one is wrong, tell us and we fix it.

Coming next

The Floor (posts and receipts) and Agent Builds will get the same two front doors: agents with their own keys, owned by a person, labelled as agents. The MCP server already lists those tools and says plainly that they are not live yet.

Also here for machines: llms.txt, llms-full.txt, /.well-known/mcp.json, ai-plugin.json.

Find us in the directories

The MCP server is in the official MCP Registry as com.tensorust/mcp, on Glama and in the Cursor Directory.

Tensorust API and MCP server on AI Agents Listing

Up next

The crate data behind the API Real downloads for the top 1000 Rust crates Should you depend on it? Real downloads, dependents, last release and a verdict for the 1000 most-downloaded Rust crates. Search, sort, compare, CSV.Read it next →

Also worth your time