INDIVIDUAL MCP TOOL

search_evidence

Free search of bounded public StackBench benchmark and failure evidence; use this for evidence-only requests.

computesage.comnone authenticationAvailability not checked

LIVE ENDPOINT

https://mcp.computesage.com/mcp

No auth detected

Connect to this endpoint to inspect the live schema for search_evidence and invoke it with your own arguments.

{}

Inferred read-only · medium confidence · heuristic, not a guarantee.

No write-capable action terms were found; this is not proof that invocation has no side effects.

CONNECT WITH APPROVAL

Review this server and its permissions before adding it. Secret placeholders must be set locally.

Codex

~/.codex/config.toml

[mcp_servers.com-computesage-stackbench]
url = "https://mcp.computesage.com/mcp"
enabled = true

Claude Code

.mcp.json

{
  "mcpServers": {
    "com-computesage-stackbench": {
      "type": "http",
      "url": "https://mcp.computesage.com/mcp"
    }
  }
}

Claude Desktop

Settings → Connectors → Add custom connector

Name: com-computesage-stackbench
Remote MCP URL: https://mcp.computesage.com/mcp

Add this remote URL as a custom connector in Claude Desktop. Availability depends on the user plan and workspace policy.

Cursor

.cursor/mcp.json

{
  "mcpServers": {
    "com-computesage-stackbench": {
      "url": "https://mcp.computesage.com/mcp"
    }
  }
}

Visual Studio Code

.vscode/mcp.json

Add to Visual Studio Code

{
  "servers": {
    "com-computesage-stackbench": {
      "type": "http",
      "url": "https://mcp.computesage.com/mcp"
    }
  }
}

Generic MCP

Client-specific MCP configuration

{
  "name": "com-computesage-stackbench",
  "transport": "streamable-http",
  "url": "https://mcp.computesage.com/mcp"
}

MCP Inspector

Run the official MCP Inspector locally and enter the indexed Streamable HTTP endpoint.

check_deployment_fit — Paid USDC Base mainnet check against existing StackBench can-it-run evidence.
predict_performance — Paid configuration performance prediction from existing StackBench evidence; returns measured-supported speed, memory, and power metrics with confidence, missing metrics, evidence, and caveats.
recommend_deployment — Paid deployment recommendation for what to buy or run: one supported configuration plus bounded alternatives, performance, confidence, evidence, caveats, and economics only when supported.
generate_launch_config — Paid deterministic vLLM or llama.