Introduction

AGENTSECURE.md

A small Markdown file generated by agentsecure start that tells humans and AI coding agents how to use the local vault and MCP broker in a repo.

The goal is simple: agents can read placeholders like ${API_KEY}, while real values stay local and are resolved only for approved API calls.

Open template

The file

Keep the generated guidance in the repo root. Do not paste real credentials, raw `.env` contents, private keys, customer data, prompts, or request bodies into this file.

# AGENTSECURE.md

## Start

agentsecure start

## What agents see

OPENAI_API_KEY=${OPENAI_API_KEY}
STRIPE_SECRET_KEY=${STRIPE_SECRET_KEY}

## Codex MCP

codex mcp add agentsecure -- agentsecure --config ./agentsecure.json mcp serve

Local vault

Import .env

Real values move into the local AgentSecure vault.

Backup

The original file is saved so humans can restore it.

Rewrite

The repo keeps safe placeholders like ${API_KEY}.

Guide

AGENTS.md explains how agents request secret-backed calls.

MCP broker

Agents should use the MCP broker for API calls that require a secret, not ask users to paste values into prompts.

tool: agentsecure.http.request
url: https://api.example.com/v1/resource
auth: ${API_KEY}
purpose: short reason for the request

Approved destinations

AgentSecure resolves placeholders from the local vault only for destinations approved by the developer or team.

approved_destinations:
  api.openai.com:
    secrets:
      - OPENAI_API_KEY
  api.stripe.com:
    secrets:
      - STRIPE_SECRET_KEY

Audit without values

Logs can show the placeholder, destination, MCP tool, and whether a request was sent. They should not include real secret values, raw .env files, prompts, request bodies, private keys, or customer data.

{
  "tool": "agentsecure.http.request",
  "placeholder": "STRIPE_SECRET_KEY",
  "destination": "api.stripe.com",
  "value_logged": false
}