AI DevKit - The Local Control Plane for AI Coding Agents<br>Skip to main contentThe future is many AI coding agents<br>Your AI coding agents are a team now. Give them a way to work together.<br>AI DevKit is the local control plane for Claude Code, Codex, Cursor, Gemini CLI, opencode, and Pi: one shared config, one console, local memory, cross-agent communication, and proof before done.<br>Built for developers already operating more than one coding agent.<br>npx ai-devkit@latest initCopy<br>Explore the docs

Works with the agents you already use

Configure the coding agents you already use. Keep the operating model consistent as your stack changes.

Before and after<br>One assistant became many agents. Now they need a way to work together.

Scattered terminal tabs<br>One local console

Duplicated agent rules<br>One reconciled config

Copy-pasted logs<br>agent send with stdin

Forgotten project decisions<br>Local memory retrieval

Done because the agent stopped<br>Done with verification evidence

The new problem is orchestration<br>AI DevKit makes them operate as one local system.

Multi-agent coding is powerful, but without shared setup, memory, communication, and verification, it becomes operationally messy. AI DevKit sits above the tools instead of replacing them.

.ai-devkit.json<br>One setup for the whole agent team<br>Reconcile Claude Code, Codex, Cursor, Gemini CLI, opencode, Pi, and more from one project-local source of truth.

agent console<br>See every running agent<br>See what is running, inspect local sessions, and keep long-running agents out of scattered terminal tabs.

agent send<br>Send work instead of copy-pasting context<br>Send prompts, logs, test output, and review tasks to the right running agent or saved agent group.

local SQLite<br>Give agents memory without bloating prompts<br>Store conventions, decisions, and reusable fixes locally so agents search when context is actually needed.

verify<br>Make done require proof<br>Use workflow skills and verification gates so completion claims need fresh build or test output.

Start in 30 seconds<br>Create your agent control plane.<br>The init flow writes project-local files you can review and commit. Re-run it whenever your agent list, skills, or workflow changes.<br>Start with one commandView on GitHub

npx ai-devkit@latest init --built-in<br>ai-devkit agent console<br>ai-devkit agent list<br>ai-devkit agent send "review this branch for release risk" --group reviewers<br>npm test 2>&1 | ai-devkit agent send --id codex --stdin<br>ai-devkit memory search --query "testing convention"