ComputeSage server
ComputeSage StackBench
Public metadata for the ComputeSage StackBench remote Model Context Protocol (MCP) server. This repository does not contain the proprietary StackBench server implementation.
Connect
Remote MCP: https://mcp.computesage.com/mcp
Website: https://computesage.com
Documentation: https://computesage.com/mcp
Official Registry:
com.computesage/stackbench
Related MCP server: ppb-mcp
Tools
search_evidencecheck_deployment_fitpredict_performancerecommend_deploymentgenerate_launch_config
Metadata
server.json mirrors the public server identity and remote
transport published in the Official MCP Registry. Service availability,
current tool schemas, and tool behavior are defined by the live remote MCP
server and its documentation.
Repository scope
This repository contains public discovery metadata only. It intentionally contains no StackBench application source, infrastructure configuration, credentials, private benchmark data, or deployment secrets.
Available Tools
5 toolscheck_deployment_fitBRead-onlyIdempotent
$0.01 USDC on Base mainnet: check model/configuration fit against bounded StackBench capacity evidence.
Inputs: model (required), context_length (required), hardware_system (required), engine, +3 more. Returns: object
generate_launch_configARead-onlyIdempotent
$0.50 USDC: generate deterministic vLLM or llama.cpp launch templates from an exact StackBench recommendation without claiming a launch occurred.
Inputs: model (required), runtime (required), prompt_tokens (required), hardware_system (required), +15 more. Returns: object
predict_performanceBRead-onlyIdempotent
$0.05 USDC: predict configuration throughput, memory, and power from bounded StackBench evidence with confidence, missing metrics, and caveats.
Inputs: model (required), prompt_tokens (required), hardware_system (required), engine, +11 more. Returns: object
recommend_deploymentCRead-onlyIdempotent
$0.25 USDC: recommend what to buy or run from bounded StackBench fit, performance, confidence, evidence, caveats, and economics only when supported.
Inputs: model (required), prompt_tokens (required), mode, engine, +38 more. Returns: object
search_evidenceARead-onlyIdempotent
$0.00 free search of bounded public StackBench benchmark and failure evidence; use this for evidence-only requests.
Inputs: query (required), limit. Returns: object
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 5 tools
Each tool targets a distinct stage in the evidence-to-deployment workflow. check_deployment_fit and predict_performance are related but clearly separated by fit versus metrics, and recommend_deployment is the only economic recommendation tool.
All five tools use consistent lowercase snake_case verb_noun names with concrete action verbs (search, check, predict, recommend, generate), making the naming pattern predictable and readable.
Five tools is appropriate for a specialized advice-and-configuration server. Every tool serves a distinct step in the workflow without redundancy or unnecessary surface area.
The server covers the full advisory pipeline: evidence search, fit validation, performance prediction, recommendation, and launch configuration generation. It intentionally stops before executing deployment, which is consistent with its stated purpose.
Maintenance
Related MCP Connectors
GPU and LLM inference benchmarks, hardware evidence, deployment recommendations, and launch configs.
Live GPU rental and LLM inference prices across dozens of providers, plus GPU price forecasts.
Measured AI-inference-storage benchmarks with citations, article search, KV-cache ROI estimation.
Compare LLM API prices, search models and providers, and access reviewed benchmark results.
Related MCP Servers
- AlicenseAqualityBmaintenanceGlobal price benchmarking for AI inference across 2,600+ SKUs from 47 vendors. Query live pricing, market indexes, and model specs via 8 tools. Free tier available.8128 npmMIT
- AlicenseNot gradedqualityDmaintenanceExposes queryable GPU inference benchmark data (quantization, throughput, VRAM, concurrent users) as tools for LLM clients.MIT
- AlicenseAqualityBmaintenanceLLM deployment planner: given a model and a GPU, answers will it fit, will it hit your SLO, and what will it cost. Sizes VRAM and KV-cache from the model's real architecture, and labels every number measured, estimated, or unknown.5979 PyPI3MIT
- AlicenseNot gradedqualityCmaintenanceEnables agents to pull measured local-LLM benchmark data and hardware-aware model picks — ranked configs with decode tok/s, quant and backend details, hardware catalogues, and repo submissions for review. Answers questions like "what should I run on a Strix Halo for coding?" with sourced numbers rather than invented ones.77 npmMIT