Configure claude-delegator with Codex MCP server
Configures Codex as specialized expert subagents via MCP for code review, architecture, and security analysis.
/plugin marketplace add jarrodwatts/claude-delegator/plugin install jarrodwatts-claude-delegator@jarrodwatts/claude-delegatorConfigure Codex (GPT) as specialized expert subagents via native MCP. Five domain experts that can advise OR implement.
which codex 2>/dev/null && codex --version 2>&1 | head -1 || echo "CODEX_MISSING"
Tell user:
Codex CLI not found.
Install with: npm install -g @openai/codex
Then authenticate: codex login
After installation, re-run /claude-delegator:setup
STOP here if Codex is not installed.
cat ~/.claude/settings.json 2>/dev/null || echo "{}"
Merge into ~/.claude/settings.json:
{
"mcpServers": {
"codex": {
"type": "stdio",
"command": "codex",
"args": ["-m", "gpt-5.2-codex", "mcp-server"]
}
}
}
Note: Use gpt-5.2-codex explicitly for the latest model.
CRITICAL:
mcpServers entriesmkdir -p ~/.claude/rules/delegator && cp ${CLAUDE_PLUGIN_ROOT}/rules/*.md ~/.claude/rules/delegator/
Run these checks and report results:
# Check 1: Codex CLI version
codex --version 2>&1 | head -1
# Check 2: MCP server configured
cat ~/.claude/settings.json | jq -r '.mcpServers.codex.args | join(" ")' 2>/dev/null
# Check 3: Rules installed (count files)
ls ~/.claude/rules/delegator/*.md 2>/dev/null | wc -l
# Check 4: Auth status (check if logged in)
codex login status 2>&1 | head -1 || echo "Run 'codex login' to authenticate"
Display actual values from the checks above:
claude-delegator Status
───────────────────────────────────────────────────
Codex CLI: ✓ [version from check 1]
Model: ✓ gpt-5.2-codex (or ✗ if not configured)
MCP Config: ✓ ~/.claude/settings.json (or ✗ if missing)
Rules: ✓ [N] files in ~/.claude/rules/delegator/
Auth: [status from check 4]
───────────────────────────────────────────────────
If any check fails, report the specific issue and how to fix it.
Setup complete!
Next steps:
1. Restart Claude Code to load MCP server
2. Authenticate: Run `codex login` in terminal (if not already done)
Five GPT experts available:
┌──────────────────┬─────────────────────────────────────────────┐
│ Architect │ "How should I structure this service?" │
│ │ "What are the tradeoffs of Redis vs X?" │
│ │ → System design, architecture decisions │
├──────────────────┼─────────────────────────────────────────────┤
│ Plan Reviewer │ "Review this migration plan" │
│ │ "Is this implementation plan complete?" │
│ │ → Plan validation before execution │
├──────────────────┼─────────────────────────────────────────────┤
│ Scope Analyst │ "Clarify the scope of this feature" │
│ │ "What am I missing in these requirements?" │
│ │ → Pre-planning, catches ambiguities │
├──────────────────┼─────────────────────────────────────────────┤
│ Code Reviewer │ "Review this PR" │
│ │ "Find issues in this implementation" │
│ │ → Code quality, bugs, maintainability │
├──────────────────┼─────────────────────────────────────────────┤
│ Security Analyst │ "Is this authentication flow secure?" │
│ │ "Harden this endpoint" │
│ │ → Vulnerabilities, threat modeling │
└──────────────────┴─────────────────────────────────────────────┘
Every expert can advise (read-only) OR implement (write).
Expert is auto-detected based on your request.
Explicit: "Ask GPT to review..." or "Have GPT fix..."
Use AskUserQuestion to ask the user if they'd like to ⭐ star the claude-delegator repository on GitHub to support the project.
Options: "Yes, star the repo" / "No thanks"
If yes: Check if gh CLI is available and run:
gh api -X PUT /user/starred/jarrodwatts/claude-delegator
If gh is not available or the command fails, provide the manual link:
https://github.com/jarrodwatts/claude-delegator
If no: Thank them and complete setup without starring.