# axme-code MCP server

> Provides persistent memory, architectural decision enforcement, and pre-execution safety hooks for AI coding agents.

- Repository: https://github.com/AxmeAI/axme-code
- Page: https://hostedmcp.org/servers/axmeai/axme-code
- Transport: —
- Stars: 14
- License: MIT

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `AXME_TELEMETRY_DISABLED` | no | no |  |
| `DO_NOT_TRACK` | no | no |  |

## FAQ

### How does AXME Code prevent dangerous commands?

It uses pre-execution safety hooks that intercept tool calls at the harness level before they execute, providing hard enforcement rather than just prompt-based suggestions.

### Can I use AXME Code with Cursor?

Yes, there is a dedicated Cursor extension that bundles the binary and registers the MCP server programmatically.

### What kind of information does the persistent knowledge base store?

It stores project structure (Oracle), architectural decisions, feedback from mistakes (Memory), safety rules, task backlogs, and session handoffs.

### Does it work with other MCP clients besides Claude Code?

Yes, it is a stdio MCP server that works with any MCP-compatible assistant like Windsurf, Cline, or Claude Desktop, though safety hooks may require specific setup.
