GPU 算力雷达 MCP server

• Registry namelink.rentgpu/gpu-marketverified in the MCP registry

The registry name sits under rentgpu.link. The registry only lets a publisher use that namespace after proving control of the domain through DNS or HTTP.

• Stateanswerschecked 2026-10-29
• Tools7
• Server calls itselfrentgpu-gpu-market
• Server version1.30.0
• Protocol version2025-06-18

On 2026-10-29 the server answered initialize and tools/list without a sign-in. First answer in 516 ms.

• Its tools do compare hourly GPU rental prices across providers; pull GPU price data via API or MCP, but not set alerts when GPU prices change.
• Checked on 4 of the 7 days it takes to judge reliability.

Said by rules from the measurements above. The likely first call is a model's pick from the tool list; read its description before relying on it.

• INITE MCP scoreafter a weeka score needs 7 daily checks: 4 so far
• What an agent can do with it2 of 3 use cases
• Answeredmeasuring4 of the 7 daily checks it takes to print a share
• Median first answer560 ms
• Space in an agent's context≈ 534 tokensthe definitions of all its tools (7), which an agent carries on every turn
• Tool definitionsC · 74 of 100
• Typed results2 of 7 toolsdeclared outputSchema: results come typed, read once

• Parameters without a description: in 4 of 7
• Not saying whether it only reads or can destroy: 7 of 7

A program checks the server once a day: initialize and tools/list, no model. Tokens are the definitions' length at about four characters a token. The grade averages six checks of every tool's name, description, schema and annotations; it says how well the tools are described, not whether they work. The score weighs what was measured and nothing else: answering (25%), doing the product's job (25%), definitions (20%), context (10%), use (10%) and how readily a client connects (10%). A part that was not measured is left out, and a hidden instruction caps the score at 40. The state of vendors' MCP servers: method and numbers →

• ping

  健康检查，免费。返回服务状态、收款地址、网络与价目表。

• list_gpu_models

  列出所有可查询的 GPU 卡型（免费）。付费查询前先调这个拿准确名称。

• gpu_price

  查某张 GPU 现在的全平台最低价（元/小时），含报价平台、30 天最低/最高/均价、当前价与 30 天最低价的差距。示例: gpu_price(gpu='RTX 4090') → 当前最低 1.24 元/小时（算家云）、市场中位 1.99、30 天最低 1.24、今日 8 家报价。按次计费 $0.002 USDC。

• gpu_history

  查某张 GPU 过去 30 天每日价格序列：每日全平台最低价、当日市场中位价、最低价来自哪个平台、库存状态、当日报价家数。示例: gpu_history(gpu='H100 80GB', limit=7) → 2026-10-20 最低 8.28 元（paperspace）、中位 23.04、有货、9 家报价。按次计费 $0.02 USDC。

• market_movers

  近 30 天 GPU 行情涨跌榜：按区间涨跌幅排序的降价/涨价卡型，含当前价与 30 天最低价。示例: market_movers(top=3) → 降价榜 A6000 48GB 2.52 元(-28.6%)、A100 80GB 5.47 元(-21.6%)、RTX 3090 0.80 元(-19.2%)。按次计费 $0.005 USDC。

• platform_reliability

  各 GPU 租用平台的数据监测覆盖度评分（A~E）：自动抓取成功率、收录天数、库存充足率、覆盖卡型数。注意：衡量的是本数据源对该平台的价格监测覆盖度，不代表平台自身的服务质量。示例: platform_reliability() → 26 家平台评分，如 gpushare 98 分 A 级（抓取成功率 97%、覆盖 7 张卡、28/29 天）。按次计费 $0.005 USDC。

• recommend_for_task

  按任务推荐最便宜的 GPU 租用方案：输入任务（如'微调 7B'/'视频生成'，或英文 finetune-7b/video-gen），返回推荐卡型、最便宜平台、每小时单价与估算任务总成本。覆盖 8 类任务：微调、推理、绘图、视频、渲染、学生练手、70B 训练、多卡集群。示例: recommend_for_task(task='微调 7B') → 最便宜卡 + 平台 + 单价 × 8 小时的总成本。这是别处没有的'按任务问价'能力。按次计费 $0.01 USDC。