What is verifyax-mcp CLAUDE.md?

A set of project instructions for VerifyAX, a platform that evaluates AI agents, plus the MCP server and typed TypeScript client that connect to it. The project is organized as a monorepo, meaning multiple related packages are kept in one repository.

Use it when developing the VerifyAX SDK or MCP server. It covers the package split, available tool scope, API authentication, local stdio operation, and remote Streamable HTTP deployment.

It records the project's scope and decisions so contributors do not redesign parts that have already been settled. It clarifies how the reusable API client and conversational MCP tools relate to each other.

verifyax-mcp CLAUDE.md — Instructions file for Claude Code by verifyax: A set of project instructions for VerifyAX, a platform…

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 instructions/verifyax/verifyax-mcp/claude-md

Clone the repo

git clone --depth 1 https://github.com/verifyax/verifyax-mcp

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

README.md

[![agentmods](https://agentmods.dev/badge/instructions/verifyax/verifyax-mcp/claude-md.svg)](https://agentmods.dev/instructions/verifyax/verifyax-mcp/claude-md)

Your own site

<a href="https://agentmods.dev/instructions/verifyax/verifyax-mcp/claude-md"><img src="https://agentmods.dev/badge/instructions/verifyax/verifyax-mcp/claude-md.svg" alt="Measured on agentmods" height="20"></a>

Per session 3,217 This file is loaded in full into every session.

When invoked 3,217 The same file — it is already loaded in full.

Security scan A 0 findings. Scan, not verified.

Origin original No closer match found in the catalogue.

Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

Model	Per session	Once invoked
Fable 5.1	$0.03217	$0.03217
Opus 5	$0.01608	$0.01608
Sonnet 5	$0.00643	$0.00643
Haiku 4.5	$0.00322	$0.00322

Measured 5d ago against content hash 2d2d77613b1e, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-06, from the pricing page.

Security

Grade A, and why

verifyax-mcp CLAUDE.md 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 5d ago.

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.

CLAUDE.md · 202 lines

How it starts

The opening of the file, as written. The whole thing — 202 lines — stays where its author put it; the contents beside it link to each section on GitHub.

CLAUDE.md

Persistent project context for Claude Code. Read this first in every session.

What we're building

A Model Context Protocol (MCP) server that exposes the VerifyAX agent-evaluation platform as conversational tools, distributed as an npm package. Plus the typed SDK it sits on top of.

Two packages in one monorepo:

@verifyax/sdk — typed TypeScript client for the VerifyAX REST API. Reusable by any consumer, not MCP-specific.
@verifyax/mcp-server — MCP server exposing ~12 tools mapped to user intents. Depends on the SDK.

The MCP server complements (does not replace) the existing verifyax-api skill at https://github.com/verifyax/verifyax-plugins. The skill is for developers writing code; the MCP server is for conversational workflows.

v1 scope and constraints

Update (0.2.0): the constraints below describe the v1 (0.1.x) shape. As of 0.2.0 the server also ships a Streamable HTTP transport + GCP Cloud Run deployment (verifyax-mcp-server-http), with per-request header auth — i.e. remote hosting, previously a v2 item, has landed. The stdio + VERIFYAX_API_KEY path below is unchanged and remains the default.

Local distribution only. Users install via npm and configure the server in their MCP client (Claude Code, Claude Desktop). Remote hosting is v2.
API key auth. User pastes their VerifyAX key into MCP client config as VERIFYAX_API_KEY. OAuth is v2.
Blocking-by-default tools. Async polling lives inside the server, invisible to Claude and the user. No start_job / get_status pairs in v1.
Structured errors over exceptions. Tools return { success: false, reason, suggested_fix } instead of throwing where possible.
Twelve tools. Resist the urge to wrap every endpoint. See the catalogue below.

Architecture decisions (already made — do not re-litigate)

Language: TypeScript. Faster-moving MCP ecosystem, better tool descriptions support.
Package manager: pnpm with workspaces.
SDK and MCP server are separate packages, published under the @verifyax npm scope.
API client style: resource-oriented, like stripe-node. client.agents.create(...), client.scenarios.generate(...), client.simulations.simulate(...).
Error handling: typed error hierarchy. VerifyaxError base; AuthError, NotFoundError, ConflictError, JobFailedError, TimeoutError, RateLimitError derived.
MCP tool errors: structured. SDK throws; the tool handler catches and translates to MCP tool-result { success: false, ... } payloads. Never let raw exceptions escape into MCP output.
Polling: in-SDK helper. client.jobs.pollUntilTerminal(jobUuid, { timeoutMs, intervalMs }). Used by both SDK consumers and the MCP server.
Logging: stderr only, structured JSON, controlled by VERIFYAX_MCP_LOG_LEVEL env. stdout is reserved for MCP protocol.
No telemetry in v1. Document this explicitly in the README.

Read the full file on GitHub · 202 lines

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.

5d ago First seen · 202 lines · 3,217 tokens per session scan A 2d2d77613b1e

Subscribe to this mod's changes

verifyax-mcp CLAUDE.md is an instructions file published in the GitHub repository verifyax/verifyax-mcp (1 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 3,217 tokens to every session, about $0.0161 per session on Opus 5. 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.

Building agents agent-evaluation agent-testing ai-agents ai-evaluation anthropic claude

verifyax-mcp CLAUDE.md

Getting it into your agent

Wrote this? Show the measurements

What it costs to keep this loaded

Grade A, and why

Nothing flagged

How it starts

CLAUDE.md

What we're building

v1 scope and constraints

Architecture decisions (already made — do not re-litigate)

What this file has done since we first saw it

Other instructions, from other repositories

nasde-toolkit CLAUDE.md

agent-triage CLAUDE.md

agent-triage copilot-instructions.md

any-agent AGENTS.md

eval-view AGENTS.md

AgentKernelArena AGENTS.md