IDE Integration

IDE Integration

AgentCTX integrates with code editors through the Model Context Protocol (MCP). This gives your AI coding assistants direct access to CTX operations — memory, knowledge search, tool routing, and trust verification.

VS Code / Cursor

Configuration

Add AgentCTX as an MCP server in your editor’s MCP settings:

{
  "mcpServers": {
    "agentctx": {
      "command": "actx",
      "args": ["start", "--protocol", "stdio"],
      "cwd": "${workspaceFolder}"
    }
  }
}

What Agents Can Do

Once connected, AI assistants in your IDE can:

Operation	CTX	What It Does
Search tools	?t github	Discover available MCP tools
Call tools	>t github.issues.create title="Fix"	Execute tools through the gateway
Search knowledge	?k "auth" #code ^3	Find relevant documentation
Store decisions	+m "choice" #arch "Rationale"	Persist architectural decisions
Verify actions	actx verify	Audit agent operations

Extension Protocol

The --protocol stdio flag enables the structured extension protocol, which uses newline-delimited JSON messages:

• ClientMessage — sent by the editor (CTX operations, configuration)
• ServerMessage — sent by AgentCTX (results, events, diagnostics)

This is different from raw MCP stdio transport — it adds IDE-specific message types for editor integration.

Any MCP-Compatible Editor

Any editor or tool that supports MCP can connect to AgentCTX:

1. Point the MCP server config to actx start
2. Use stdio transport (default)
3. The actx tool (~300 tokens) appears as a single callable

Docker Environment

If your team runs AgentCTX via Docker Compose, configure the editor to connect to the SSE endpoint:

{
  "mcpServers": {
    "agentctx": {
      "url": "http://localhost:3100",
      "transport": "sse"
    }
  }
}

See Also

• actx start — transport options
• Quick Start — initial setup