iris

A local MCP server for scoring coding-agent outputs. It evaluates each result for quality, safety, and cost using the Iris evaluation package.

In plain words
What is it for?
Adding output evaluation to an agent setup, starting the Iris MCP server, and scoring generated results for quality, safety, and cost.
Why use it?
It gives teams a repeatable way to assess agent results instead of judging them only by intuition. The server runs locally and is configured with a log-level setting.

MCP server for Claude CodeCodexCursorGemini CLIOpenCode

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add mcp/iris-eval/mcp-server/iris
Clone the repo
git clone --depth 1 https://github.com/iris-eval/mcp-server

Made for: Claude Code, Codex, Cursor, Gemini CLI, OpenCode.

Per session not measured What this adds to a session before it is invoked.
When invoked not measured Not applicable: nothing here is loaded into a session.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Security

Grade A, and why

iris scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured yesterday.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.mcp.json · 12 lines

What it actually says

{
  "iris": {
    "command": "npx",
    "args": [
      "-y",
      "@iris-eval/mcp-server"
    ],
    "env": {
      "IRIS_LOG_LEVEL": "info"
    }
  }
}
Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. yesterday First seen · 12 lines scan A 14fa195ed15c

Subscribe to this mod's changes

iris is an MCP server published in the GitHub repository iris-eval/mcp-server (7 stars, last pushed 8d ago), licensed MIT. Its token cost is not measured: an MCP server costs its tool schemas, not its config file. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other mcp servers, from other repositories

evalview

MCP server "evalview" as configured in hidai25/eval-view. Launched with evalview mcp serve. Needs 2 environment variables to run.

hidai25/eval-view · not measured

evalview-mcp

Open-source testing and regression detection framework for AI agents. Golden baseline diffing, CI/CD integration, works with LangGraph, CrewAI, OpenAI, Anthropic Claude, HuggingFace, Ollama, and MCP. Runs locally from the evalview Python package. Needs 1 environment variable to run.

hidai25/eval-view · not measured

governance-mcp

MCP server "governance-mcp" as configured in cirwel/unitares. Runs locally from the governance-mcp Python package.

cirwel/unitares · not measured

rule-drift

Test whether AI agents retain critical instructions as conversations grow. Runs locally from the rule-drift npm package.

ri7in/rule-drift · not measured

memtrust

Independent, reproducible benchmark harness for agent-memory backends (MemPalace, Mem0, Zep/Graphiti, OpenViking): runs LongMemEval, LoCoMo, contradiction-detection, and a dozen other evals against all four and publishes the raw logs, not vendor-curated numbers. Runs locally from the memtrust-cli Python package.

RudrenduPaul/memtrust · not measured

deskcert

Certify whether an AI agent is safe to operate your internal web application before production rollout. Author a YAML task suite against your own app, run it with Playwright, and gate CI/CD on a forbidden-action safety score. Runs locally from the deskcert-cli Python package.

RudrenduPaul/DeskCert-CLI · not measured