Claude Code Directive MCP
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@Claude Code Directive MCP Enable Claude Code directive until next Friday"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Claude Code Directive MCP
claude-code-directive-mcp is a local stdio MCP server that lets an agent temporarily add a global Codex AGENTS directive requiring Claude Code usage until a configured expiry date.
It is designed for temporary, opt-in delegation policy:
Adds one clearly marked block to the effective global Codex instruction file.
Uses Codex precedence: non-empty
AGENTS.override.md, otherwiseAGENTS.md.Skips project-local
.codexdirectories unless explicitly opted in.Records a manifest and creates a user-level cleanup timer.
Removes only the exact marked block recorded in the manifest.
Adds guardrails for Claude context, Main Codex context, retries, parallelism, and rate limits.
Never asks for or stores secrets.
Install
git clone <repo-url>
cd claude-code-directive-mcp
npm install
npm run install:codexRestart Codex after installation so the new MCP server is loaded.
The installer writes a user wrapper to:
~/.local/bin/codex-claude-code-directive-mcpand registers this MCP server in:
${CODEX_HOME:-~/.codex}/config.tomlRelated MCP server: codex-bridge
MCP Tools
claude_directive_enable
Enable or update the temporary directive.
Example arguments:
{
"expiresAt": "2026-08-17T23:59:59",
"timezone": "Europe/Berlin",
"cleanupDelayMinutes": 5,
"maxClaudePromptChars": 60000,
"maxClaudeOutputChars": 20000,
"maxClaudeRuntimeSeconds": 900,
"maxClaudeRetries": 1,
"mainContextReservePercent": 25,
"maxClaudeCallsPerTask": 3,
"maxParallelClaudeAgents": 2,
"minSecondsBetweenClaudeCalls": 30
}claude_directive_status
Reports discovered Codex homes, effective global instruction files, marker counts, scheduler state, manifest content, and optional Claude CLI status.
claude_directive_disable
Runs the idempotent cleanup and disables/removes the one-time user scheduler.
Cleanup
On Linux with a working user systemd, the MCP creates a user timer named:
codex-claude-code-directive-cleanup.timerThe timer executes:
${CODEX_HOME:-~/.codex}/maintenance/cleanup-temporary-claude-code-directive.pyThe cleanup script uses the manifest to remove only the exact marked directive block. It is safe to run repeatedly.
Budget And Rate Guardrails
The directive generated by this MCP tells the main agent to:
keep a Main Codex context reserve,
cap Claude prompt and output size,
cap Claude runtime and retries,
limit per-task Claude calls and parallel Claude agents,
wait between separate Claude calls unless parallelism is justified,
stop fast retry loops on rate limits, context limits, timeouts, or overload.
The MCP writes these rules into the global instruction block; it does not have privileged access to live provider token counters.
Available Tools
3 toolsclaude_directive_disableDisable Temporary Claude Code DirectiveB
Remove the active temporary Claude Code directive block using the manifest and disable the one-time cleanup scheduler.
Inputs: dryRun.
claude_directive_enableEnable Temporary Claude Code DirectiveB
Enable or update a global temporary Codex AGENTS directive that tells Codex to use Claude Code until a specific expiry, with context guardrails and scheduled cleanup.
Inputs: expiresAt (required), dryRun, timezone, maxClaudeRetries, +9 more.
claude_directive_statusTemporary Claude Code Directive StatusC
Report discovered Codex homes, effective AGENTS files, current directive markers, scheduler state, context policy, and optional Claude CLI availability.
Inputs: runClaudeTest, includeClaudeCheck.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 3 tools
Each tool has a clearly distinct function: enabling, checking status, or disabling the directive. No overlap in purposes.
All tools follow a consistent `claude_directive_` prefix with verb_noun pattern (enable, status, disable), making them easily identifiable.
Three tools is exactly right for managing a lifecycle (create/read/delete) of a single directive; compact and well-scoped.
The set covers the full lifecycle: enabling (create/update), checking status (read), and disabling (delete). No obvious gaps for the stated domain.
Related MCP Connectors
Paid remote MCP for Claude Code skill update gate MCP, structured receipts, audit logs, and reviewer
A paid remote MCP for OpenAI Codex agent coordination MCP, built to return verdicts, receipts, usage
Shared memory MCP server for Claude Code, Codex, Gemini CLI and Cursor. Cited, editable.
Set up the JEV model router for Claude Code, Codex and Cursor, and install MCP servers.
Related MCP Servers
- AlicenseAqualityBmaintenanceMCP server that enables Claude Code to send code and plans for review by OpenAI Codex, returning structured feedback inline.5420 npm27MIT
- AlicenseAqualityDmaintenanceMCP server that lets Claude Code ask GPT Codex for adversarial planning, code review, debugging, research, and risk triage without leaving your project workflow.99 npm2MIT
- AlicenseNot gradedqualityBmaintenanceMCP server enabling Claude to consult Codex (GPT-5.x) mid-task for second opinions, plan/diff review, brainstorming, and codebase exploration via structured debates and permission-controlled interactions.2MIT
- AlicenseBqualityDmaintenanceMCP server that enables Claude Code to delegate code generation and debugging to OpenAI Codex, with stall detection and auto-recovery.214 npmMIT