Update to Compute Screener

Changed files

Choose a file to see its text changes. Binary or large files show metadata only.

plugin.json →

plugin.json

− removed · + added

--- before
+++ after
@@ -1,8 +1,8 @@
 {
   "$schema": "https://agent-plugins.org/schemas/1.0.0/plugin.schema.json",
   "name": "compute-screener",
-  "version": "1.0.2",
-  "description": "Daily price indices for renting GPUs and for AI model tokens, with history, provider averages and exchange prices.",
+  "version": "1.4.0",
+  "description": "Compute prices: GPU rental price indices, compute futures prices and AI token prices, with history.",
   "author": {
     "name": "Compute Screener",
     "email": "support@computescreener.com",
@@ -10,11 +10,12 @@
   },
   "homepage": "https://computescreener.com/connect",
   "keywords": [
+    "compute prices",
+    "compute futures",
     "gpu",
     "h100",
     "b200",
     "gpu rental",
-    "compute",
     "ai tokens",
     "llm pricing",
     "price index"
@@ -23,14 +24,19 @@
     "com.openai": {
       "interface": {
         "displayName": "Compute Screener",
-        "shortDescription": "GPU and AI token prices",
-        "longDescription": "Compute Screener tracks what AI compute costs.\n\nGPU rental prices: a daily price index in USD per GPU-hour for H100, H200, A100, B200, B300 and RTX PRO 6000, with 1-day, 1-week and 1-month changes. Each index is built from the posted on-demand prices of cloud providers and GPU marketplaces, standardized to one benchmark contract.\n\nProviders and exchanges: for one GPU, each provider's average price, and the last price and 24-hour volume of the contracts on that GPU's rental price at the exchanges that list them.\n\nAI token prices: a daily index in USD per 1M tokens for 10 leading AI model makers, with each company's models, their share of usage and their posted prices.\n\nPrice history: daily values for any GPU or company, up to 1,000 days back.\n\nEvery result links to the GPU's or company's page on computescreener.com, and the methodology is public at computescreener.com/methodology. No account is needed: the plugin only reads published prices and keeps no record of requests. Prices are information, not investment advice.",
+        "shortDescription": "GPU rental and AI token prices",
+        "longDescription": "Compute prices: daily GPU rental indices for H100, H200, A100, B200, B300, RTX PRO 6000, L40S, RTX A6000, RTX 6000 Ada, L40 and RTX 4090, compute futures prices on the exchanges that list them, and AI token prices for 13 AI companies.\n\nGPU rental prices: a daily price index in USD per GPU-hour for each GPU the site lists, with 1-day, 1-week and 1-month changes. Each index is recalculated every minute from the posted on-demand prices of cloud providers and GPU marketplaces (20 sources), standardized to one benchmark contract.\n\nCompute futures and providers: for one GPU, each exchange's compute futures price (the last price and 24-hour volume of the contracts on that GPU's rental price) and each provider's average price. Every contract is at computescreener.com/futures.\n\nAI token prices: a daily index in USD per 1M tokens for 13 AI model makers: each company's models at their posted prices for a cache-heavy mix (87% cache reads, 12% input, 1% output), weighted by usage, with each model's share of usage and posted prices.\n\nPrice history: daily values for any GPU or company, up to 1,000 days back, and any GPU's price at a past date or time.\n\nModels and providers: each AI model's posted API prices and price changes, each GPU provider's posted prices, and every other GPU's posted prices (RTX 5090, A40, L4 and over 70 more).\n\nChanges and checks: what moved over 1, 5, 7 or 30 days, the break-even throughput of each model on each GPU, and for any final value the manifest of the files it was computed from, so anyone can recompute it.\n\nMore data: wholesale electricity prices at Texas and California hubs, AI coding tools by usage, and AI data center sites, through search and fetch.\n\nEvery answer is what the website shows and links to its page on computescreener.com; the methodology is public at computescreener.com/methodology. No account is needed: the plugin only reads published prices and keeps no record of requests.",
         "developerName": "Compute Screener",
         "category": "Data & Analytics",
         "capabilities": [
           "GPU rental prices",
+          "Compute futures prices",
           "AI token prices",
-          "Price history"
+          "Price history",
+          "AI model API prices",
+          "GPU provider prices",
+          "Price checks",
+          "Break-even throughput"
         ],
         "websiteURL": "https://computescreener.com",
         "supportURL": "https://computescreener.com/support",
@@ -38,7 +44,7 @@
         "termsOfServiceURL": "https://computescreener.com/terms",
         "defaultPrompt": [
           "What does it cost to rent an H100 per hour?",
-          "Compare the rental prices of H200, B200 and B300.",
+          "What are H100 compute futures trading at?",
           "Which AI company's tokens cost the most?"
         ],
         "brandColor": "#1A73E8",
@@ -70,7 +76,7 @@
               "description": "Compare AI companies' token prices",
               "prompt": "Which AI company's tokens cost the most right now?",
               "tools_triggered": "get_token_prices",
-              "expected_behavior": "Ranks the listed companies by their token index in USD per 1M tokens (3 input : 1 output blend) and names the most expensive one."
+              "expected_behavior": "Ranks the listed companies by their token index in USD per 1M tokens (each model's posted prices at 87% cache reads, 12% input and 1% output, weighted by usage) and names the most expensive one."
             },
             {
               "description": "Price history over a period",
@@ -99,7 +105,7 @@
         "commerce_description": "The plugin shows public price data. It does not sell products, process payments or place orders."
       },
       "publication": {
-        "release_notes": "First release: GPU rental price index, AI token price index, price history, provider averages and exchange prices of GPU rental contracts."
+        "release_notes": "Every answer holds only what a page of computescreener.com shows: the collectors' source addresses, a model's 3:1 list price and figures the Every AI model page does not list were removed; a price at a past time is the chart's own point. Answers what computescreener.com shows, and only that. GPU rental indices for eleven GPUs (L40S, RTX A6000, RTX 6000 Ada, L40 and RTX 4090 added). Token Price Index 2.0: posted prices at a cache-heavy mix (87% cache reads, 12% input, 1% output) for 13 companies, with its history recomputed from 2025. Each AI model's and GPU provider's posted prices, every other GPU's posted prices, price changes, break-even throughput and checks of final values. Search and fetch also read the site's pages on wholesale electricity prices, AI coding tools by usage and AI data center sites. Tools for figures the site does not show (cost estimates, ranges and volatility, market overview, token volume, GPU supply, futures curve, model fit) were removed."
       },
       "apps": "./.app.json"
     }