Skip to main content
Glama

Plures

Audit the agent's past decisions

audit_my_agent

Sends a Plures audit: up to 50 decisions an AI agent already made, graded so the developer sees which were wrong. The account's first audit is free. After it has finished, each further audit is paid from the account's credit when it is sent: $0.20 a decision, or $0.50 for kind judgment, so tell the developer the cost and get their yes before sending a second one. One audit at a time. Works with a test or live key. Each row is input (what the agent saw, including the policy it should follow) and decision (what it did). Gather them from the developer's logs, database or test fixtures; leave out names and other private details (emails, phones and card numbers are removed anyway). kind check (default) asks whether each call was right (the dashboard calls it the accuracy audit); kind judgment asks whether a customer would be happy with it (the quality audit). Text only: to have an image, video or audio clip judged, the developer sends a quality audit from plures.si/dashboard/audit with a media column. A person at Plures reviews it before anyone grades it. Use get_audit_report to read the results.

Tool call template

This tool declares no required parameters. Add optional parameters from the input schema as needed.

{
  "id": 1,
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "arguments": {},
    "name": "audit_my_agent"
  }
}

Input Schema

NameRequiredDescriptionDefault
csvNoInstead of rows: CSV text with a header row and columns input and decision
kindNocheck (default): was each call right. judgment: would a customer be happy with it
rowsNoThe decisions: what the agent saw and what it did
aboutNoWhat the agent does, in a sentence, e.g. "Answers support tickets and can refund up to $50"

Schema Changelog

Changes observed during successful MCP inspections.

  1. First observed

TDQS

A4.5/5.0
Behavior4/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Adds substantial behavior beyond the annotations (readOnly=false, openWorld=true, idempotent=false): the first audit is free, later audits bill $0.20/decision or $0.50 for judgment, charges occur at send time, the developer must be told the cost and consent, and a human at Plures reviews media audits. The one gap is that it never says whether a submitted audit can be cancelled or what happens to a partially processed batch.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness3/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Purpose and cost are front-loaded well, but the middle is a run-on chain where pricing, key types, row format, PII stripping, the kind mapping and media routing all run together in unbroken sentences. Nearly every clause is load-bearing, yet splitting it into short labelled blocks would make the consent and media-exclusion rules far easier to scan.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

For a 4-parameter submission tool with no output schema, the description covers what to submit, where to get it, what gets scrubbed, what it costs, when consent is required, the free-first-audit rule, and explicitly redirects to get_audit_report for results. Nothing an agent needs to call it correctly is missing.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters4/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

With 100% schema coverage the baseline is 3, but the description adds real meaning: rows are paired input (what the agent saw, including the governing policy) and decision (what it did), sourced from logs, databases or test fixtures, with PII like emails, phones and card numbers removed. It maps the kind enum to the dashboard names (accuracy audit vs quality audit), which the schema does not do; only about and csv get no extra gloss.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

States a specific verb and resource: it sends a Plures audit of up to 50 already-made decisions and has them graded so the developer sees which were wrong. It separates itself from the reporting sibling by ending with 'Use get_audit_report to read the results', so an agent can distinguish submitting an audit from fetching one.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Gives explicit routing: default kind check is the accuracy audit, kind judgment is the quality audit, and media grading is explicitly NOT this tool – it goes through plures.si/dashboard/audit. It also states the sequencing rule (one audit at a time), the consent rule before a paid second audit, and names get_audit_report for follow-up.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Try in Browser

Glama MCP Gateway

Add one secure layer between your agents and this server.