Numguard MCP Server

Verify a number before an agent asserts it — evals, backtests (Deflated Sharpe), signed receipts.

Remote serverstreamable-httpPython

What is the Numguard MCP server?

Numguard MCP server is a hosted integration for AI assistants that speak the Model Context Protocol. Verify a number before an agent asserts it — evals, backtests (Deflated Sharpe), signed receipts.

What you get

MCP registry identity — mcp-name: io.github.ipezygj/numguard

What the assistant can call

Once Numguard is connected, these are the calls the assistant has available:

  • verify_subset_winDoes "we lead on subset X" survive correcting for how many subsets I tested?
  • verify_model_gapIs the gap between these two models bigger than the test set can resolve?
  • verify_judge_biasIs my judge's preference real, or just longer / first / same-family?
  • calibrate_judgeIs the LLM judge I trust actually calibrated against ground truth?
  • audit_leaderboardIs #1 on this leaderboard statistically real? (rank confidence intervals)
  • verify_agentThis wallet claims a track record — fetch its own public on-chain trades and re-derive the result.
  • audit_addressesRun that same verdict across an explicit list of addresses. (only the addresses given)
  • verify_vaultRe-derive a vault's APY from its own Deposit/Withdraw events, instead of quoting its page.
  • verify_backingRe-derive backing = reserves held / token supply, from the chain.
  • verify_guard_traceRecompute a behavioural-guard verdict over an agent's action trace, and sign it.
  • check_attestationLook up a numguard credential on-chain. (free, no key, no gas)
  • erc8004_feedbackBuild the ERC-8004 giveFeedback call from a verdict, so reputation carries the evidence. (free)

Setting it up

Being a remote server, there is no local install. You register the endpoint with your client, authorise it once, and the tools appear.

Configuration and credentials

You will need one environment variable: NUMGUARD_FACILITATOR_URL. The server will not start without them, which is usually why the tools fail to appear on a first run. Keep credentials in your client's env block or a secrets manager rather than in a file you might commit.

Choosing this one

Among the AI and media services options, the useful question is rarely "what can it do" but "what does it cost you to run" — permissions, credentials, and how much of your context its toolset consumes. Numguard's toolset — verify_subset_win, verify_model_gap, verify_judge_bias and 9 more — is a fair guide to whether it matches your workflow. It is maintained by ipezygj; worth a glance at recent repository activity before you build anything load-bearing on it.

SyncDev reviews every entry in this directory against the project's own documentation before publishing, and revisits them as servers change.

Before you rely on it

  • Your data travels to the provider's service, so the usual questions apply about what you send and what they retain.
  • With 12 tools registered it takes up a noticeable share of the context window; turn it off in projects that never touch Numguard.
  • Missing credentials fail quietly in some clients — if no tools show up, check the environment block first.
  • MCP clients confirm each tool call by default. Leave that on until you have watched what the numguard mcp server does with a few real requests.

Available tools

ToolWhat it does
verify_subset_win*Does "we lead on subset X" survive correcting for how many subsets I tested?*
verify_model_gap*Is the gap between these two models bigger than the test set can resolve?*
verify_judge_bias*Is my judge's preference real, or just longer / first / same-family?*
calibrate_judge*Is the LLM judge I trust actually calibrated against ground truth?*
audit_leaderboard*Is #1 on this leaderboard statistically real?* (rank confidence intervals)
verify_agent*This wallet claims a track record — fetch its own public on-chain trades and re-derive the result.*
audit_addresses*Run that same verdict across an explicit list of addresses.* (only the addresses given)
verify_vault*Re-derive a vault's APY from its own Deposit/Withdraw events, instead of quoting its page.*
verify_backing*Re-derive backing = reserves held / token supply, from the chain.*
verify_guard_trace*Recompute a behavioural-guard verdict over an agent's action trace, and sign it.*
check_attestation*Look up a numguard credential on-chain.* (**free**, no key, no gas)
erc8004_feedback*Build the ERC-8004 giveFeedback call from a verdict, so reputation carries the evidence.* (**free**)

How to install the Numguard MCP server

{
  "mcpServers": {
    "numguard": {
      "command": "uvx",
      "args": ["numguard"],
      "env": {
        "NUMGUARD_FACILITATOR_URL": "your-value"
      }
    }
  }
}

Add to claude_desktop_config.json, then restart Claude Desktop.

Configuration

VariableDescriptionRequired
NUMGUARD_FACILITATOR_URLEndpoint or connection string the server talks to.Yes

Example prompts to try

  • Use Numguard to verify subset win.
  • Use Numguard to verify model gap.
  • Use Numguard to verify judge bias.

Frequently asked questions

It connects Numguard to MCP-compatible AI assistants such as Claude and Cursor, exposing 12 tools (verify_subset_win, verify_model_gap, verify_judge_bias, and more) that the assistant can call on your behalf. Instead of copying data back and forth by hand, the assistant works with Numguard directly.