Iris MCP Server

Iris scores every agent run for quality, safety, and cost - on your machine, with no SDK and no account. Most agent projects check quality by running a few remembered prompts and eyeballing the output. Iris replaces that with numbers you can audit: your agent's runs land in a SQLite database on your disk, 25 built-in rules score them deterministically - PII, prompt injection, hallucination markers, cost thresholds…

People who work with agent evaluation, eval and observability and want it reachable from Claude, Cursor, VS Code, or another MCP client. The project is written in TypeScript.

VERIFIED ACTIVE

LAST COMMIT 2026-09-30 · ★ 9 · #16 OF 125 MAINTAINED DEVOPS & MONITORING · VERIFIED 2026-09-30

MIT · TypeScript servers · how we verify → /methodology

01 · Install Iris

before you install - you'll need

Set IRIS_ANTHROPIC_API_KEY before connecting. IRIS_API_KEY is optional or environment-specific per the README.

Claude Code

claude mcp add iris-eval-mcp-server -- npx -y @iris-eval/mcp-server

Claude Desktop / Cursor / VS Code - add to config

{
  "mcpServers": {
    "iris-eval-mcp-server": {
      "command": "npx",
      "args": [
        "-y",
        "@iris-eval/mcp-server"
      ]
    }
  }
}

Same JSON for Cursor. For VS Code, rename the top-level key from `mcpServers` to `servers`.

Using another client? Same JSON, different key

Claude Desktop · mcpServers

Cursor · mcpServers

VS Code · servers

Windsurf · mcpServers

Zed · context_servers

Cline · mcpServers

Roo Code · mcpServers

Continue · mcpServers

LibreChat · mcpServers

Gemini CLI · mcpServers

Codex CLI · mcp_servers

Full setup guides: every client.

02 · Evidence

Security posture

What to check before giving this server access to your agent - from the registry, GitHub, and our own probes. We don't score safety; we show what's verifiable.

runs as local process (stdio) - runs on your machine with your user's permissions

license MIT - declared in the repository

npm package @iris-eval/mcp-server - published under the repo owner's npm scope (@iris-eval)

registry namespace io.github.iris-eval is GitHub-verified and matches the repo owner

03 · What Iris can do

Prose above is summarized from the project's README and registry record - no invented capabilities.

Latest releases

v0.19.0 · 2026-09-25

[0.19.0] - 2026-09-25 · Verdicts that say how sure they are, detectors that see through disguises, and one command to set up any client. 0.19.0: · A verdict is called decisive only where labelled data shows its risk…

v0.18.0 · 2026-09-24

[0.18.0] - 2026-09-24 · Honest numbers, a lighter footprint, nothing done silently. 0.18.0 finishes the work of the 0.17.0 security review: · The published accuracy numbers no longer count the real transcripts they are…

v0.17.0 · 2026-09-23

[0.17.0] - 2026-09-23 · Hostile input, honest verdicts, a harder release path. An adversarial review of 0.16.0 attacked Iris the way a determined user or a compromised dependency would. 0.17.0 closes what it found…

04 · Who maintains Iris

Iris is maintained by iris-eval. It's the only MCP server we track from this author; the repo dates to Mar 2026.

05 · Facts

repository: github.com/iris-eval/mcp-server
website: https://iris-eval.com/
category: DevOps & monitoring - ranked #16 of 125 actively-maintained DevOps & monitoring servers as of 2026-09-30.
release cadence: 10+ releases in the last 90 days (latest 2026-09-25)
registry: io.github.iris-eval/mcp-server (active, first published 2026-03-15 · 33 versions)
packages: npm:@iris-eval/mcp-server

06 · Iris FAQ

What is Iris?

Is Iris still maintained?

Yes - as of 2026-09-30, its last commit was 2026-09-30 and it shipped 10+ releases in the last 90 days. We re-verify nightly.

How do I install Iris?

Run `npx -y @iris-eval/mcp-server`. The README documents 2 environment variables (IRIS_ANTHROPIC_API_KEY, IRIS_API_KEY) to set first. Set IRIS_ANTHROPIC_API_KEY before connecting. IRIS_API_KEY is optional or environment-specific per the README. You can also paste the ready-made client config above.

Does Iris run locally?

Yes - it's a stdio server: it runs on your machine (via npx) with your user's permissions. Your data stays local unless the server itself calls external APIs.

07 · Alternatives to Iris

Alternatives to Iris

Maintained DevOps & monitoring servers if Iris isn't the fit.

Pairs well with

Servers that cover what Iris doesn't - only shown when the pairing reason fits the companion.

More DevOps & monitoring MCP servers · Data Sentry · Forge Craft · Thoughtbox · Mushi Mushi

More TypeScript MCP servers · Stellaroid Earn · MERN MCP · Forage · A11y Check · Kolada MCP Server · see all