Agent configurations and capabilities change frequently. If you find information on this page that is out of date or incorrect, please create a pull request or open an issue to help us keep this documentation current.

MCP support

AlignTrue supports Model Context Protocol (MCP) configuration for 15 AI coding agents via 14 MCP exporters. MCP enables AI agents to connect to external tools, data sources, and services through a standardized protocol.

What is MCP?

Model Context Protocol allows AI agents to:

Access external tools and resources
Integrate with project-specific services
Extend agent capabilities beyond training data
Connect to custom APIs and data sources

AlignTrue automatically generates agent-specific MCP configuration files from your rules, ensuring consistent tool access across your development workflow.

Supported MCP agents

AlignTrue exports MCP configuration via 14 MCP exporters covering 15 agents across three format families:

JSON (11 exporters): Most agents, including VS Code, Cursor, Amazon Q, and Roo Code
YAML (2 exporters): Goose and Trae AI
Directory-structured (1 exporter): Junie uses a nested MCP layout

JSON-based MCP configurations

Agent	Exporter	Config File	Notes
GitHub Copilot	`vscode-mcp`	`.vscode/mcp.json`	VS Code integration
Cursor	`cursor-mcp`	`.cursor/mcp.json`	Cursor-specific setup
Claude Code	`root-mcp`	`.mcp.json`	Shared root format (also used by Aider)
Aider	`root-mcp`	`.mcp.json`	Shared root format (also used by Claude)
Windsurf	`windsurf-mcp`	`.windsurf/mcp_config.json`	Windsurf-specific format
Amazon Q	`amazonq-mcp`	`.amazonq/mcp.json`	AWS integration
Firebase Studio	`firebase-mcp`	`.idx/mcp.json`	Firebase IDX platform
KiloCode	`kilocode-mcp`	`.kilocode/mcp.json`	KiloCode-specific
Roo Code	`roocode-mcp`	`.roo/mcp.json`	Roo Code format
Amp	`amp-mcp`	`.amp/settings.json`	Project-level settings
AugmentCode	`augmentcode-mcp`	`.augment/settings.json`	Augment platform
Kiro	`kiro-mcp`	`.kiro/settings/mcp.json`	Kiro settings directory

YAML-based MCP configurations

Agent	Exporter	Config File	Notes
Goose	`goose-mcp`	`.goose/config.yaml`	YAML format for extensions
Trae AI	`traeai-mcp`	`trae_config.yaml`	Project root (git-ignored)

Special MCP handling

Agent	Exporter	Config File	Notes
Junie	`junie-mcp`	`.junie/mcp/mcp.json`	Directory-based MCP structure

Central MCP configuration

AlignTrue uses a centralized MCP configuration in .aligntrue/config.yaml:


mcp:
  servers:
    - name: my-custom-server
      command: python
      args: ["./tools/custom-mcp.py"]
      env:
        API_KEY: "${process.env.CUSTOM_API_KEY}"
    - name: nodejs-server
      command: node
      args: ["./mcp-server.js"]

This central definition is propagated to agent-specific MCP configuration files automatically during sync.

Quick start: generate MCP configs

Define servers once in .aligntrue/config.yaml under mcp.servers (see example above).
Run aligntrue sync from the repo root to generate all agent configs deterministically.
Open outputs in agent locations:
- .vscode/mcp.json (GitHub Copilot)
- .cursor/mcp.json (Cursor)
- .mcp.json (root format for Claude Code/Aider)
- Agent-specific paths like .windsurf/mcp_config.json, .amazonq/mcp.json, .roo/mcp.json, .kiro/settings/mcp.json
- For Trae AI, trae_config.yaml lives at the repo root—add it to .gitignore if it should remain local-only.

Exporter capabilities

Each MCP exporter:

Reads MCP server definitions from .aligntrue/config.yaml
Generates agent-specific MCP server configuration files
Propagates your centralized server definitions to each agent’s required format
Includes content hash for determinism and verification
Supports vendor-specific customizations

Configuration schema

MCP servers require:

name (required): Unique identifier (e.g., aligntrue)
command (required): Executable command (e.g., npx, python, node)
args (optional): Command arguments as array
env (optional): Environment variables as object
disabled (optional): Temporarily disable without removing config

Format differences

JSON vs YAML

JSON-based (most agents):

Consistent structure across platforms
Easier parsing and validation
Direct MCP protocol representation

YAML-based (Goose, Trae):

Human-readable format
Easier manual editing
Agent-specific customizations

Project-level only

By design, AlignTrue writes MCP configuration only to project-level files:

.vscode/mcp.json (not global VS Code settings)
.cursor/mcp.json (not global Cursor config)
.goose/config.yaml (not global ~/.config/goose/config.yaml)

This approach:

Keeps rules version-controlled and reviewable
Enables team collaboration on shared MCP config
Prevents machine-specific drift
Works with monorepos and scoped rules

If you need global MCP configuration, copy the project-level file to your global config location manually.

Fidelity notes

When MCP features cannot be fully represented, exporters include fidelity notes:


{
  "fidelity_notes": [
    "Section 'Performance': machine-checkable checks not represented in MCP format"
  ]
}

These notes indicate:

Information loss during export
Unsupported features for this agent
Recommendations for manual customization

Determinism

All MCP exports are deterministic:

Same input (rules + config) → identical MCP JSON
Content hash included for verification
Order preserved across exports
Reproducible across machines and CI

Agent Support - Full compatibility matrix
AI Agent Guide - Quick reference for all agents
Sync Behavior - How MCP configs sync
MCP Protocol - Official MCP specification