Multi-Platform Setup Guide
September 6, 2026 · View on GitHub
Agentic QE supports 11 coding agent platforms via MCP (Model Context Protocol). This guide covers setup for all platforms.
Quick Setup
The fastest way to configure any platform:
# Auto-setup with init flag
npx agentic-qe init --auto --with-copilot
npx agentic-qe init --auto --with-cursor
npx agentic-qe init --auto --with-all-platforms
# Or use the platform command directly
npx agentic-qe platform setup copilot
npx agentic-qe platform setup cursor
# List all platforms and their status
npx agentic-qe platform list
# Verify a platform's configuration
npx agentic-qe platform verify copilot
Platform Reference
| Platform | Config Path | Format | Rules File | Auto-Detect |
|---|---|---|---|---|
| Claude Code | .claude/mcp.json | JSON | CLAUDE.md | Built-in |
| OpenCode | opencode.json | JSON | — | opencode.json |
| Kiro | .kiro/settings/mcp.json | JSON | .kiro/steering/ | .kiro/ |
| GitHub Copilot | .vscode/mcp.json | JSON | .github/copilot-instructions.md | .vscode/ |
| Cursor | .cursor/mcp.json | JSON | .cursorrules | .cursor/ |
| Cline | .vscode/cline_mcp_settings.json | JSON | .vscode/cline_custom_modes.json | — |
| Kilo Code | .kilocode/mcp.json | JSON | .kilocode/modes.json | .kilocode/ |
| Roo Code | .roo/mcp.json | JSON | .roo/modes.json | .roo/ |
| OpenAI Codex CLI | .codex/config.toml | TOML | AGENTS.md | .codex/ |
| Windsurf | .windsurf/mcp_config.json | JSON | .windsurfrules | .windsurf/ |
| Continue.dev | .continue/config.yaml | YAML | .continue/rules/aqe-qe-standards.yaml | .continue/ |
GitHub Copilot
Automated Setup
npx agentic-qe init --auto --with-copilot
# or
npx agentic-qe platform setup copilot
Manual Setup
Create .vscode/mcp.json:
{
"servers": {
"agentic-qe": {
"type": "stdio",
"command": "npx",
"args": ["-y", "agentic-qe@latest", "mcp"],
"env": {
"AQE_MEMORY_PATH": ".agentic-qe/memory.db",
"AQE_V3_MODE": "true"
}
}
}
}
The installer also creates .github/copilot-instructions.md with QE behavioral rules.
Note: Copilot uses
serversas the config key, notmcpServers.
Cursor
Automated Setup
npx agentic-qe init --auto --with-cursor
# or
npx agentic-qe platform setup cursor
Manual Setup
Create .cursor/mcp.json:
{
"mcpServers": {
"agentic-qe": {
"type": "stdio",
"command": "npx",
"args": ["-y", "agentic-qe@latest", "mcp"],
"env": {
"AQE_MEMORY_PATH": ".agentic-qe/memory.db",
"AQE_V3_MODE": "true"
}
}
}
}
The installer also creates .cursorrules with QE behavioral rules.
Cline
Automated Setup
npx agentic-qe init --auto --with-cline
# or
npx agentic-qe platform setup cline
Manual Setup
Create .vscode/cline_mcp_settings.json:
{
"mcpServers": {
"agentic-qe": {
"command": "npx",
"args": ["-y", "agentic-qe@latest", "mcp"],
"env": {
"AQE_MEMORY_PATH": ".agentic-qe/memory.db",
"AQE_V3_MODE": "true"
},
"alwaysAllow": [
"fleet_init", "fleet_status", "fleet_health",
"agent_list", "agent_metrics", "agent_status",
"team_list", "team_health",
"task_list", "task_status",
"test_generate_enhanced",
"coverage_analyze_sublinear", "quality_assess",
"defect_predict", "code_index",
"memory_store", "memory_retrieve", "memory_query", "memory_usage",
"model_route", "routing_metrics", "aqe_health"
]
}
}
}
The installer also creates a custom QE Engineer mode in .vscode/cline_custom_modes.json.
Cline-specific: Supports
alwaysAllowfor auto-approving safe read-only tools.
Kilo Code
Automated Setup
npx agentic-qe init --auto --with-kilocode
# or
npx agentic-qe platform setup kilocode
Manual Setup
Create .kilocode/mcp.json:
{
"mcpServers": {
"agentic-qe": {
"command": "npx",
"args": ["-y", "agentic-qe@latest", "mcp"],
"env": {
"AQE_MEMORY_PATH": ".agentic-qe/memory.db",
"AQE_V3_MODE": "true"
},
"alwaysAllow": [
"fleet_init", "fleet_status", "fleet_health",
"agent_list", "agent_metrics", "agent_status",
"team_list", "team_health", "task_list", "task_status",
"test_generate_enhanced", "coverage_analyze_sublinear",
"quality_assess", "defect_predict", "code_index",
"memory_store", "memory_retrieve", "memory_query", "memory_usage",
"model_route", "routing_metrics", "aqe_health"
]
}
}
}
The installer also creates a custom QE mode in .kilocode/modes.json.
Roo Code
Automated Setup
npx agentic-qe init --auto --with-roocode
# or
npx agentic-qe platform setup roocode
Manual Setup
Create .roo/mcp.json:
{
"mcpServers": {
"agentic-qe": {
"command": "npx",
"args": ["-y", "agentic-qe@latest", "mcp"],
"env": {
"AQE_MEMORY_PATH": ".agentic-qe/memory.db",
"AQE_V3_MODE": "true"
},
"alwaysAllow": [
"fleet_init", "fleet_status", "fleet_health",
"agent_list", "agent_metrics", "agent_status",
"team_list", "team_health", "task_list", "task_status",
"test_generate_enhanced", "coverage_analyze_sublinear",
"quality_assess", "defect_predict", "code_index",
"memory_store", "memory_retrieve", "memory_query", "memory_usage",
"model_route", "routing_metrics", "aqe_health"
]
}
}
}
The installer also creates a custom QE mode in .roo/modes.json.
OpenAI Codex CLI
Automated Setup
npx agentic-qe init --auto --with-codex
# or
npx agentic-qe platform setup codex
Codex guidance is independent from MCP, hooks, and skills. Choose the eager
AGENTS.md payload with --codex-guidance full|compact|none on either command.
full is the default. compact keeps only the safety and verification
contract and is capped at 512 UTF-8 bytes including AQE's ownership sentinels
(a conservative ceiling of 171 planning tokens at three bytes per token).
none removes only well-formed AQE-owned sentinel blocks. It leaves all other
AGENTS.md bytes untouched and still provisions MCP, hooks, and skills.
npx agentic-qe init --auto --with-codex --codex-guidance compact
npx agentic-qe platform setup codex --codex-guidance none
npx agentic-qe platform verify codex --codex-guidance none
Ruflo integration is deliberately opt-in. Add it only when you want its development-time coordination guidance and lifecycle hooks:
npx agentic-qe init --auto --with-codex --with-ruflo
# or
npx agentic-qe platform setup codex --with-ruflo
Without --with-ruflo, setup prints that Ruflo was not installed and shows the
opt-in flag. AQE hooks and the five default Codex QE skills are still installed.
Manual Setup
Create .codex/config.toml:
[mcp_servers.agentic-qe]
type = "stdio"
command = "npx"
args = ["-y", "agentic-qe@latest", "mcp"]
[mcp_servers.agentic-qe.env]
AQE_MEMORY_PATH = ".agentic-qe/memory.db"
AQE_V3_MODE = "true"
The installer provisions the complete repository-scoped Codex QE surface:
.codex/config.tomlwith the Agentic QE MCP server;AGENTS.mdwith QE behavioral rules;.codex/hooks.jsonand local AQE hook adapters/runtimes for routing, guardrails, and outcome learning;- curated Codex-native QE skills under
.agents/skills/.
When --with-ruflo is selected, setup additionally installs the Ruflo adapter,
runtime bridge, lifecycle groups, and aqe-ruflo guidance skill. It does not
add Ruflo as an Agentic QE runtime dependency.
Verify every installed component after setup:
npx agentic-qe platform verify codex
For deliberate variants, verify the same intent explicitly:
npx agentic-qe platform verify codex --no-mcp
npx agentic-qe platform verify codex --with-ruflo
The command reports MCP config, instructions, hooks, adapters, runtimes, and discoverable QE skills separately, and exits unsuccessfully when a component is missing or invalid.
Codex-specific: Uses TOML config format and
AGENTS.mdfor behavioral instructions.
Windsurf
Automated Setup
npx agentic-qe init --auto --with-windsurf
# or
npx agentic-qe platform setup windsurf
Manual Setup
Create .windsurf/mcp_config.json:
{
"mcpServers": {
"agentic-qe": {
"command": "npx",
"args": ["-y", "agentic-qe@latest", "mcp"],
"env": {
"AQE_MEMORY_PATH": ".agentic-qe/memory.db",
"AQE_V3_MODE": "true"
}
}
}
}
The installer also creates .windsurfrules with QE behavioral rules.
Continue.dev
Automated Setup
npx agentic-qe init --auto --with-continuedev
# or
npx agentic-qe platform setup continuedev
Manual Setup
Create .continue/config.yaml:
mcpServers:
- name: agentic-qe
command: npx
args:
- "-y"
- "agentic-qe@latest"
- "mcp"
env:
AQE_MEMORY_PATH: ".agentic-qe/memory.db"
AQE_V3_MODE: "true"
The installer also creates .continue/rules/aqe-qe-standards.yaml with QE rules in YAML format.
Continue.dev-specific: Uses YAML config format with array-style MCP server entries.
Multi-Platform Setup
To configure all platforms at once:
npx agentic-qe init --auto --with-all-platforms
This creates config files for all 8 platforms. Files are merge-safe — existing configs are preserved and AQE entries are appended.
Auto-Detection
When running npx agentic-qe init --auto, platforms are auto-detected based on directory presence:
| Directory Exists | Platform Auto-Configured |
|---|---|
.vscode/ | GitHub Copilot |
.cursor/ | Cursor |
.kilocode/ | Kilo Code |
.roo/ | Roo Code |
.codex/ | OpenAI Codex CLI |
.windsurf/ | Windsurf |
.continue/ | Continue.dev |
Note: Cline is not auto-detected (shares
.vscode/with Copilot). Use--with-clineexplicitly.
Verifying Configuration
# Check all platforms
npx agentic-qe platform list
# Verify a specific platform
npx agentic-qe platform verify cursor
Verify checks:
- Config file exists at expected path
- Config file is valid JSON/TOML/YAML
- Config contains
agentic-qeserver entry - Rules/behavioral file exists
Troubleshooting
Config not detected: Run npx agentic-qe platform verify <name> to see which checks fail.
Merge conflicts: The installer merges safely into existing configs. If you see issues, delete the AQE entry from the config and re-run setup.
MCP server not starting: Ensure npx agentic-qe mcp works standalone first:
npx agentic-qe@latest mcp