Interactive CLI playground for testing AgentDock workflows.

AgentDock CLI provides a local interactive environment for testing agents, tools, sessions, and workspace operations through the AgentDock harness.

✨ Features

• Interactive Agent Sessions: Run prompts continuously in a local REPL.
• Session Persistence: Save, inspect, and resume sessions.
• Workspace Tools: Read, search, list, write, and update files inside a workspace.
• Workspace File Tools: Execute approved file operations inside the workspace.
• TypeScript First: Fully typed and built for Node.js applications.

🚀 Setup

Install the dependencies and build the CLI:

yarn install
yarn build

Create a .env file in this directory:

OPENROUTER_API_KEY=your-key
# Optional: overrides the CLI's explicit catalog default.
OPENROUTER_MODEL=your-model-id

The CLI loads this file automatically. Do not commit it.

💻 Local Development

Run the CLI directly from TypeScript:

yarn dev

Run the typecheck and production build:

yarn typecheck
yarn build

🛠️ Usage

Start an interactive AgentDock session using the current working directory as the workspace:

yarn start

yarn dev uses a local .sandbox workspace for safe testing. The installed agentdock command uses the directory where it is launched as the workspace.

Once running, enter prompts continuously. Use /help for interactive commands. Use /model or /models to browse and select a model; the selected model is shown in the header.

Sessions are stored under .agentdock/sessions/ in the directory where the CLI is launched.

Resume a saved session from the command line:

yarn dev --resume <session-id>
agentdock --resume <session-id>

Inside the CLI, use /resume to list saved sessions and /resume <session-id> to switch to one.

🧾 Debug Logging

Readable, colorized logs are written to logs/agentdock-cli.log when logging is enabled. Entries include their source module, such as main or agent:

INFO  [main] agentdock-cli starting workspace=/Volumes/Code/Github/my-project
DEBUG [agent] tool started tool=list_files
INFO  [agent] agent prompt completed durationMs=8312 chunks=17 tools=1

AGENTDOCK_LOGGING=true
tail -f logs/agentdock-cli.log

Set AGENTDOCK_LOGGING=false to disable file logging.

Reset the active log file when starting a fresh debugging session:

yarn logs:clear