Generic MCP Setup

Configure ACE MCP with any MCP-compatible AI tool or application.

Prerequisites

An MCP-compatible AI tool
ACE CLI installed and logged in
ACE server running (ace start)

Standard MCP Configuration

Most MCP-compatible tools use a similar JSON configuration format. Add this to your tool's MCP config:

{
  "mcpServers": {
    "ace": {
      "command": "ace",
      "args": ["mcp", "start"],
      "env": {
        "ACE_NAMESPACE": "your-project-name"
      }
    }
  }
}

Configuration Options

ACE_NAMESPACE

The namespace isolates memories for different projects. Each project should have its own namespace.

"ACE_NAMESPACE": "my-project"

ACE_SERVER_URL

Override the default server URL (useful for remote ACE servers):

"ACE_SERVER_URL": "http://localhost:7777"

ACE_JWT_TOKEN

Provide authentication token for secured deployments:

"ACE_JWT_TOKEN": "your-jwt-token"

Manual MCP Server Start

You can also start the ACE MCP server manually for testing:

# Start ACE MCP server with a specific namespace
ACE_NAMESPACE=my-project ace mcp start

# Or with all options
ACE_NAMESPACE=my-project \
ACE_SERVER_URL=http://localhost:7777 \
ace mcp start

Testing the Connection

Once configured, test the integration by asking your AI tool to:

"Remember that this project uses [technology]" - Tests ace_remember
"What do you know about [topic]?" - Tests ace_recall
"Track an issue: [description]" - Tests ace_track_issue
"Get project statistics" - Tests ace_get_stats

Available ACE MCP Tools

Memory Operations

ace_remember - Store memories
ace_recall - Search memories
ace_get_context - Get comprehensive context
ace_search - Deep semantic search

Issue Tracking

ace_track_issue - Log bugs/problems
ace_resolve_issue - Mark as resolved
ace_find_solutions - Search past solutions

Documentation

ace_track_decision - Record decisions
ace_architecture - Document components
ace_patterns - Store patterns
ace_best_practices - Guidelines

Knowledge Graph