Interactive CLI playground for testing Agentdock workflows.

Agentdock CLI is a local playground for a LangChain createAgent backed by a LangGraph checkpointer. It consumes AgentDock events while keeping model configuration, tools, workspace access, and persistence in the CLI application.

✨ Features

• Interactive Agent Sessions: Run prompts continuously in a local REPL.
• Checkpointed Conversations: Save and resume threads through LangGraph's SQLite saver.
• Workspace Tools: Read, search, list, write, and update files inside a 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/, and LangGraph checkpoints are stored in the SQLite database .agentdock/checkpoints.sqlite in the directory where the CLI is launched. The CLI owns the SQLite saver and closes its database handle when it exits. Approval requests resume through LangGraph using the same authorized thread.

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