Configures AI coding agents like Claude Code, Cursor, OpenCode and Codex with persistent memory, shared workflows, skills and MCP servers.
codeaholicguy/ai-devkit
A local-first control plane that unifies config, session monitoring, cross-agent messaging, and memory for AI coding agents like Claude Code and Codex CLI.
- Stars
- 1,661
- Forks
- 261
- Open issues
- 37
- Last commit
- 4 hours ago
What is it?
AI DevKit is a local operating layer for teams that run several AI coding agents at once. A single .ai-devkit.json file reconciles rules, skills and MCP setup across tools such as Claude Code, Codex CLI, Gemini CLI, opencode, Cursor and GitHub Copilot, so each agent does not need hand-maintained configuration.
The project provides an agent console that lists and inspects running sessions, and an agent send command that routes prompts or piped stdin to a running session or a saved group of agents. Project decisions, conventions and fixes are stored in a local SQLite memory that agents search on demand through MCP and CLI rather than loading into every prompt.
AI DevKit also installs a set of engineering skills, anchored by dev-lifecycle, which walks an agent through requirements, design, planning, implementation, testing and review. Session control is supported on macOS and Linux through tmux, and the project is licensed under Apache-2.0.
Key features
- 01Generates per-agent config, rules and MCP setup from one .ai-devkit.json file.
- 02Live terminal console for listing and inspecting running agent sessions.
- 03Sends prompts or piped stdin to a running agent or a saved agent group.
- 04Stores and searches project memory in local SQLite via MCP and CLI.
- 05Ships built-in skills such as dev-lifecycle, verify, tdd and dev-debug.
Use cases
- Supervising several coding agents running in separate terminals from one console.
- Keeping the same project rules and skills consistent across Claude Code, Cursor and Codex CLI.
- Piping test output or logs into a running agent session for follow-up.
- Running structured requirements-to-review workflows for larger or riskier changes.
Quick start
- 1
Run machine setup once
npx ai-devkit@latest setup - 2
Enter your project directory
cd your-project - 3
Initialize AI DevKit in the project
npx ai-devkit@latest init
Things to know
- Agent console and tmux-based session control work only on macOS and Linux, not Windows.
- The agent console is marked experimental and requires tmux.
In collections
Topics
FAQ
- What is ai-devkit?
- A local-first control plane that unifies config, session monitoring, cross-agent messaging, and memory for AI coding agents like Claude Code and Codex CLI. AI DevKit is a local operating layer for teams that run several AI coding agents at once. A single .ai-devkit.json file reconciles rules, skills and MCP setup across tools such as Claude Code, Codex CLI, Gemini CLI, opencode, Cursor and GitHub Copilot, so each agent does not need hand-maintained configuration.
- Is ai-devkit free to use? What license does it have?
- ai-devkit is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
- Is ai-devkit still maintained?
- Actively maintained. The last commit to ai-devkit was on Nov 7, 2026.
- What can I use ai-devkit for?
- Supervising several coding agents running in separate terminals from one console. Keeping the same project rules and skills consistent across Claude Code, Cursor and Codex CLI. Piping test output or logs into a running agent session for follow-up. Running structured requirements-to-review workflows for larger or riskier changes.
- How do I install ai-devkit?
- Run machine setup once: npx ai-devkit@latest setup Enter your project directory: cd your-project Initialize AI DevKit in the project: npx ai-devkit@latest init
- Where can I get ai-devkit?
- The source code and installation instructions are on GitHub: https://github.com/codeaholicguy/ai-devkit
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
An open-source coding agent that runs as a desktop app, a terminal CLI, or an installable Skill for Codex, OpenCode and DeepSeek Harness.
A local catalog and control plane that lets Codex or Claude search 2,700+ SKILL.md playbooks, pick exact skills, validate a stack and preview an install plan.
A Pi-native terminal workspace that puts parallel coding agents, live tasks, Git changes, and permission controls into one interface.
A terminal UI for viewing, checking, installing and updating agent skills across Claude Code, OpenCode, Codex, Cursor and other coding agents.
A CLI that installs a project-local harness of skills, rules, hooks and knowledge docs for 18 AI coding assistants, after you review the plan.