Configuration
AGENTS.md is an open format file that tells AI coding agents how to work in your project before they plan any change.
It is read automatically by OpenAI Codex, GitHub Copilot, Cursor, Windsurf, and most other coding agents. Think of it as a project briefing that every agent reads before touching your code.
The Zephex template below tells your agents which endpoint to use, which tools exist, and the order to call them for the best results.
# Zephex MCP Configuration ## MCP Endpoint- URL: https://zephex.dev/mcp- Auth: Authorization: Bearer YOUR_API_KEY- Protocol: JSON-RPC 2.0 over Streamable HTTP- Tool count: 10 hosted tools ## Mandatory Tool OrderAlways follow this sequence for coding tasks:1. check_test — Identify the minimal files to read BEFORE reading anything2. get_project_context — Run once per session for project orientation3. read_code — Read specific functions/classes (not whole files)4. find_code — Search for patterns before writing new code5. keep_thinking — Track reasoning state for complex or high-risk tasks ## Available Tools Reference| Tool | Purpose | When to Call ||---|---|---|| check_test | Minimal file set for the task | FIRST, before any file reads || get_project_context | Project brief + full file tree | Once per session (covers structure) || read_code | AST symbol extraction + callers | Instead of reading full files || find_code | Repo-wide pattern search | Before implementing anything new || explain_architecture | Request/auth/data flow tracing | Before touching auth or DB code || check_package | npm package verification | Before adding dependencies || project_memory | Remember project decisions via local SQLite | During code review / new session || audit_headers | HTTP/TLS security audit | After every deploy || Zephex_dev_info | Expert KB (search → get) | Auth, DB, security patterns—not URL scraping || keep_thinking | Structured reasoning + session state | Complex debugging, high-risk changes | ## Rules- Never read a full file when read_code with a symbol name will do- Never implement something new before running find_code to check if it exists- Always call check_test at the start of non-trivial tasks- Call audit_headers after any deployment to production ## Project Context[Fill in your project-specific notes here: stack, key directories, naming conventions, deployment targets]Agents read AGENTS.md in this order, with the nearest file winning:
~/.agents.md for global defaults.Put the template at ./AGENTS.md for a project-specific setup, or ~/.agents.md if you want Zephex instructions to apply across all projects on the machine.
| File | Works with |
|---|---|
| CLAUDE.md | Claude Code only |
| AGENTS.md | OpenAI Codex, GitHub Copilot, Cursor, Windsurf, JetBrains, and most others |
For full coverage, create both files. They can contain identical content.
| Editor | Reads AGENTS.md | Reads CLAUDE.md |
|---|---|---|
| OpenAI Codex | ✓ | ✗ |
| GitHub Copilot | ✓ | ✗ |
| Cursor | ✓ | ✗ |
| Windsurf | ✓ | ✗ |
| JetBrains AI | ✓ | ✗ |
| Claude Code | ✗ | ✓ |