v0.1.0 — Initial Release
First public release of @donghae0414/codex-gemini-mcp.
A minimal TypeScript MCP server that lets AI agents (Claude, Cursor, etc.) call OpenAI Codex CLI and Google Gemini CLI directly as MCP tools over stdio transport.
Features
MCP Tools
ask_codex— Request code generation, refactoring, and debugging from Codex CLIask_gemini— Request analysis, summarization, and code review from Gemini CLIwait_for_job— Block until a background job completescheck_job_status— Poll the status of a background jobkill_job— Send SIGTERM/SIGINT to a running joblist_jobs— List jobs filtered by status
Background Execution
- Long-running CLI calls run as background jobs by default (
background: true) - Job state and I/O persisted under
.codex-gemini-mcp/jobs/and.codex-gemini-mcp/prompts/
Structured Logging
- JSONL logging to
.codex-gemini-mcp/logs/mcp-YYYY-MM-DD.jsonl request_id↔job_idbidirectional correlation- Optional preview/full-text logging via
MCP_LOG_PREVIEW/MCP_LOG_FULL_TEXT
Reliability
- Output cap enforcement (
MCP_MAX_OUTPUT_BYTES, default 1 MiB) - Model name regex validation
- TTY/color disabled for stable pipe output
Requirements
- Node.js 20+
codexCLI:npm i -g @openai/codexgeminiCLI:npm i -g @google/gemini-cli
Install
npm i -g @donghae0414/codex-gemini-mcpAcknowledgements
Inspired by the Codex/Gemini MCP servers originally built in [oh-my-claudecode](https://github.com/Yeachan-Heo/oh-my-claudecode).