Skip to content

MULTI_CLIENT.md

Latest commit

 

History

History
191 lines (130 loc) · 5.67 KB

MULTI_CLIENT.md

File metadata and controls

191 lines (130 loc) · 5.67 KB

Multi-Client Support

AXME Code is built as a stdio MCP server — it works with any MCP-compatible AI coding assistant, not just Claude Code.

Compatibility Matrix

Client MCP Tools Safety Hooks Session-end Auditor File-change Tracking
Claude Code (CLI / VS Code) ✅ Full ✅ Full ✅ Automatic ✅ Automatic
Cursor ✅ Full ❌ ❌ ❌
Windsurf ✅ Full ❌ ❌ ❌
Cline (VS Code) ✅ Full ❌ ❌ ❌
Claude Desktop ✅ Full ❌ ❌ ❌
Any other MCP client ✅ Full ❌ ❌ ❌

MCP tools available everywhere — the full set of 19 tools (axme_context, axme_save_memory, axme_save_decision, axme_safety, axme_status, etc.) works in any MCP client. Your agent can read and write the knowledge base.

Hooks are Claude Code-only — pre-execution safety hooks, post-tool-use file tracking, and the session-end background auditor depend on Claude Code's hook system. In other clients, the agent must explicitly call AXME tools (e.g., to save memories at session close).

Quick Start (any client)

1. Install AXME Code

curl -fsSL https://raw.githubusercontent.com/AxmeAI/axme-code/main/install.sh | bash

Installs to ~/.local/bin/axme-code.

2. Initialize the project

cd your-project
axme-code setup

This builds the knowledge base (.axme-code/) and creates .mcp.json for Claude Code. For other clients, copy the MCP server entry into your client's config (see below).

3. Configure your client

The MCP server entry is identical for every client:

{
  "command": "axme-code",
  "args": ["serve"]
}

Just place it in the right config file for your client.

Per-Client Setup

Claude Code

Already done by axme-code setup — .mcp.json is created in your project root. Just run claude and start using it.

Cursor

Add to ~/.cursor/mcp.json (global) or .cursor/mcp.json (per-project):

{
  "mcpServers": {
    "axme": {
      "command": "axme-code",
      "args": ["serve"]
    }
  }
}

Restart Cursor. The agent will have access to all axme_* tools.

Windsurf (Codeium)

Add to ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "axme": {
      "command": "axme-code",
      "args": ["serve"]
    }
  }
}

Restart Windsurf.

Cline (VS Code extension)

Open VS Code Settings → search for "Cline MCP" → edit cline_mcp_settings.json:

{
  "mcpServers": {
    "axme": {
      "command": "axme-code",
      "args": ["serve"]
    }
  }
}

Reload the VS Code window.

Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or equivalent:

{
  "mcpServers": {
    "axme": {
      "command": "axme-code",
      "args": ["serve"]
    }
  }
}

Restart Claude Desktop.

Generic MCP Client

Any MCP client that supports stdio transport will work. The server is launched with:

axme-code serve

It reads from stdin and writes to stdout following the MCP JSON-RPC protocol.

What You Get Without Hooks

In non-Claude-Code clients, you don't get automatic hook-based features. Workarounds:

Safety enforcement

Hooks aren't available — the agent can still call axme_safety to check rules before running commands, but enforcement is advisory rather than blocking. Recommendation: rely on the agent's discipline, or wrap dangerous commands in a confirmation prompt at the application layer.

Session-end audit

The background auditor doesn't trigger automatically. Instead, ask your agent to call axme_begin_close and axme_finalize_close at the end of work — the agent walks through the same checklist (memories, decisions, safety rules) and writes the handoff.

File-change tracking

Without PostToolUse hooks, file changes aren't automatically logged. The agent can manually note significant changes in memories or via axme_worklog.

Multiple Clients on the Same Project

You can use multiple clients on the same project — they all read and write to the same .axme-code/ storage. Conflicts are unlikely because:

  • All writes go through atomic file operations (atomicWrite)
  • Session metadata is keyed by unique session IDs
  • Append-only worklog handles concurrent sessions

The knowledge base is the single source of truth regardless of which client created an entry.

Limitations

  • Hooks are not portable — Claude Code's hook system is unique. Other clients will not get automatic safety blocking, file tracking, or session-end audits.
  • CLAUDE.md auto-injection — only Claude Code reads CLAUDE.md. Other clients ignore it; you may need to provide instructions to your agent another way (e.g., a system prompt).
  • .mcp.json is Claude Code-specific — axme-code setup writes this file, but other clients use different config locations (see above).
  • Session-to-session linking — AXME tracks Claude Code's session IDs to link conversations across VS Code windows. Other clients don't expose session IDs the same way, so cross-window session linking may not work; sessions will still be tracked individually.

Verifying It Works

After configuration, ask your agent in any client:

Call axme_context to load the project knowledge base.

If you see oracle, decisions, memories, and safety rules in the response — AXME is working.

See Also