mcpproxy-go CLAUDE.md

Project instructions for mcpproxy-go, a Go service that acts as a proxy for model tools and follows autonomous working rules.

In plain words
What is it for?
They guide feature work through failing Go tests, standard formatting, mocked fallbacks for unclear dependencies, and escalation for destructive changes.
Why use it?
They set defaults for making safe assumptions, require test-first progress, and define when work must stop for human input.

Instructions file

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/smart-mcp-proxy/mcpproxy-go/claude-md
Clone the repo
git clone --depth 1 https://github.com/smart-mcp-proxy/mcpproxy-go
Per session 3,515 This file is loaded in full into every session.
When invoked 3,515 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.

ModelPer sessionOnce invoked
Fable 5 $0.03515 $0.03515
Opus 5 $0.01758 $0.01758
Sonnet 5 $0.00703 $0.00703
Haiku 4.5 $0.00351 $0.00351

Measured 2d ago against content hash 680d508e34ad, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

mcpproxy-go 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 2d 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 · 165 lines

How it starts

The opening of the file, as written. The whole thing — 165 lines — stays where its author put it; the contents beside it link to each section on GitHub.

CLAUDE.md

Guidance for Claude Code / AI agents working in this repo. This file is loaded into every session and every Paperclip heartbeat — keep it lean. It is orientation + behavior only; detailed reference lives in docs/.

Autonomous Operation Constraints

Must-Do (Defaults & Assumptions)

  • Zero Interruption Policy: If a decision is needed and no explicit instruction exists, make an informed, safe assumption based on idiomatic Go best practices and document it in the PR/commit. Do NOT ask for human clarification mid-task.
  • Test-Driven Progress: Write a failing Go test (_test.go) for every sub-task before implementing the feature.
  • Graceful Fallbacks: If an API or dependency lacks documentation, use mock interfaces or a simplified implementation rather than blocking the task.

Must-Nots

  • Do NOT ask for plan approval: Once a plan/spec is generated, begin execution immediately.
  • Do NOT stop for code style choices: Run gofmt/goimports and follow standard Go conventions.

Escalation Triggers (Stop Conditions)

Only halt and ask a human IF:

  1. You need destructive data operations or to delete core proxy logic that cannot be mocked.
  2. A required environment variable is missing from .env and cannot be mocked for the task's scope.
  3. You are stuck in an error loop for the same go test failing after 5 consecutive attempts.
  4. Cross-model (Codex) review round cap — per PR: when a PR is gated by a cross-model Codex review, run at most 5 fix→re-review rounds on that PR. If Codex has not returned a clean verdict after the 5th round, STOP and ask the human how to proceed (do not auto-run round 6). The counter is per-PR and resets for each new PR. (Verify each Codex finding is genuine before fixing — Codex can false-positive; a round only counts when you push a fix and re-review.)

Project Overview

MCPProxy is a Go desktop application that acts as a smart proxy for AI agents using the Model Context Protocol (MCP): intelligent tool discovery, massive token savings, and built-in security quarantine against malicious MCP servers.

Read the full file on GitHub · 165 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.

  1. 2d ago First seen · 165 lines · 3,515 tokens per session scan A 680d508e34ad

Subscribe to this mod's changes

mcpproxy-go CLAUDE.md is an instructions file published in the GitHub repository smart-mcp-proxy/mcpproxy-go (326 stars, last pushed 3d ago), licensed MIT. It adds 3,515 tokens to every session, about $0.0176 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 2026-08-30.