codex-omnifocus-mcp
Enables Codex to interact with the user's local OmniFocus database on macOS, allowing task management, project creation, and tag operations through natural language.
README
codex-omnifocus-mcp
Local Codex-focused MCP server for OmniFocus on macOS.
This project starts from omnifocus-mcp-enhanced and adds Codex-oriented server instructions, MCP resources, compact query tools, tag tools, clean shutdown, logging, and write audit logging.
What This Repo Is
This repository is the source code for a local MCP server. It is not a hosted web service and it does not run a central cloud server.
Each user installs this repo on their own Mac. Codex starts the MCP server locally with Node.js, talks to it over stdio, and the server reads or updates the user's local OmniFocus database through AppleScript/JXA.
In short:
Codex -> local MCP stdio -> node dist/server.js -> AppleScript/JXA -> OmniFocus
The GitHub repo's role is distribution: source code, docs, install scripts, tests, and release history.
Install From A Clone
git clone https://github.com/phd-peter/codex-omnifocus-mcp.git ~/.local/share/codex-omnifocus-mcp
cd ~/.local/share/codex-omnifocus-mcp
scripts/install-codex.sh
To clone from a script-driven install after downloading this script separately:
CODEX_OMNIFOCUS_REPO_URL=https://github.com/phd-peter/codex-omnifocus-mcp.git scripts/install-codex.sh
The installer runs npm ci, builds the TypeScript output, and registers the MCP server with Codex:
codex mcp add codex-omnifocus -- node /path/to/codex-omnifocus-mcp/dist/server.js
If Codex does not show the new tools immediately, restart Codex or open a new Codex session.
Local Development
npm install
npm run build
codex mcp add codex-omnifocus -- node "$PWD/dist/server.js"
The server is intended for local use. package.json remains marked private as an npm publishing guard; the GitHub repository can still be public.
Useful commands:
npm test
npm run build
npm run preflight
npm run register:codex
Create Your Own OmniFocus Agent
This repo provides the shared MCP server and the default Codex tool-use guidance in skills/codex-omnifocus/SKILL.md. Your personal agent behavior should live in a separate AGENTS.md so each user can define their own OmniFocus system without changing the server.
Start from the template:
cp AGENTS.example.md AGENTS.md
Then edit AGENTS.md for your own workflow:
- default capture destination
- tags and custom perspectives
- due/defer date habits
- daily planning and weekly review style
- when the agent may write directly and when it should ask first
Personal AGENTS.md files are ignored by default because they can include sensitive project names, routines, or decision rules.
Audit Logs
Write tools append JSONL entries to:
~/.codex/codex-omnifocus-mcp/audit/YYYY-MM-DD.jsonl
Override the directory with:
export CODEX_OMNIFOCUS_AUDIT_DIR=/path/to/audit
Audited tools:
add_omnifocus_taskadd_projectedit_itemmove_taskremove_itembatch_add_itemsbatch_remove_itemscreate_tag
Verification
npm test
npm run build
npm audit --omit=dev
OmniFocus integration tests should be run manually with OmniFocus open on macOS.
Privacy
See PRIVACY.md. Audit logs and OmniFocus exports can contain private task, project, tag, date, and note data. Do not commit backups/, exports/, .env*, audit/, or *.jsonl.
Architecture
See docs/architecture.md for the local runtime model and trust boundary.
Distribution
See docs/distribution.md for the public GitHub, local-first release strategy.
Marketing
See docs/marketing.md for launch positioning, channels, content ideas, and the first 30 days of public distribution work.
Notices
See NOTICE.md for upstream attribution.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。