Run only the AWS services you need, locally

What a workspace is

Goku runs real engines in containers next to it. Each Lambda function, ECS task, RDS database, OpenSearch domain, MSK cluster and Glue notebook takes memory while it runs, and on a laptop that adds up. A workspace is a saved group of the services you work with together: start it when you need them, and pause or stop it when you don't.

• “Serverless day”: Lambda, ECS and DynamoDB.
• “Data day”: Glue, S3 and Athena.
• As many as you like. Several workspaces can run at once, and one service can belong to several of them.

Each workspace shows its state (Running, Paused, Stopped, or Partial with the reason), its memory, CPU and containers, the time it was active today, and a history of who did what: the console, the CLI, MCP or auto-pause.

Start, pause, resume, stop and switch

• Shared services stay up. A service in two running workspaces is shared: pausing or stopping one of them leaves it running, and the step says “kept running: shared with” the other one.
• Core services are always on. S3, DynamoDB, Lambda, SQS, SNS, IAM, KMS, Secrets Manager, CloudWatch and Systems Manager keep answering. A workspace still pauses and stops their containers, such as those of your Lambda functions.
• Your data is kept. Stop frees memory in Goku and in Docker; buckets, tables and databases stay in the data folder.
• Every step is shown. An action lists its steps as they run, one action runs at a time per workspace, and a failed step doesn't stop the others.

Install or update Goku

Workspaces need Goku 4.2 and goku CLI 0.6.0 or later. On macOS and Linux:

$ curl -fsSL https://tanuj24.github.io/goku/install.sh | sh

On Windows, in PowerShell:

PS> irm https://tanuj24.github.io/goku/install.ps1 | iex

Already installed? Run goku update --cli, then goku update. More options are on the install page.

Create a workspace

In the console

Open Workspaces in the console's sidebar, http://localhost:8080/workspaces, and choose New workspace. Start from a template (Serverless, Data & analytics, Containers, Databases or Everything), from the services on now, or from a blank list, then pick a name, a color, the services and when to auto-pause. The Services page and the “Running outside any workspace” card also have Save as workspace.

From a terminal

$ goku ws create "Serverless day" --services lambda,ecs,dynamodb
goku ws create "Data day" --services glue,s3,athena --color teal
goku ws create today --from-current --auto-pause 30   # the services on now

Services are ids or names such as lambda, "Step Functions" or msk; goku services lists them. --template takes serverless, data, containers, databases or everything, and --color takes orange, blue, green, purple, pink, teal, amber or slate. A workspace's id is its name in lower case with dashes, such as serverless-day, and every command also takes the name, in any case.

Start, pause, stop and switch

$ goku ws start serverless-day    # turn its services on (starts Goku too if needed)
goku ws pause serverless-day    # no CPU; the containers keep their memory
goku ws resume serverless-day
goku ws stop serverless-day     # free the memory, keep the data
goku ws switch data-day         # start Data day, stop every other workspace

Each action shows its steps and the state and memory it leaves. goku ws ls shows them all:

$ goku ws ls
NAME            STATE    SERVICES               MEMORY    LAST USED
Data day        stopped  glue, s3, athena       -         3 days ago
Serverless day  running  lambda, ecs, dynamodb  412.0 MB  now

2 workspaces: 1 running, 1 stopped · 412.0 MB in use

In the console, the workspace switcher in the top bar does the same in one click, from any page.

Auto-pause, and what runs outside workspaces

Auto-pause pauses a running workspace once all of its containers have been idle for the time you pick: 15, 30 or 60 minutes, or any other time from 5 minutes to 24 hours. Pinned resources are left alone.

$ goku ws edit data-day --auto-pause 60
goku ws edit serverless-day --auto-pause off

Services and containers that belong to no running workspace are listed as Outside workspaces, with their memory and a Stop button. Starting, pausing or stopping a workspace leaves them alone. goku ws stop-outside stops them, and Save as workspace keeps them as a new one.

Share a workspace as JSON

Export a workspace and commit it next to your code, so everyone on the team starts the same services:

$ goku ws export data-day -o data-day.json
goku ws import data-day.json                   # or --name "Data night" to add it under another name

{
  "name": "Data day",
  "description": "Glue ETL into the lake, Athena on top.",
  "color": "teal",
  "services": ["glue", "s3", "athena"],
  "options": { "autoPauseMinutes": 60 }
}

The console's Import button on the Workspaces page reads the same file.

Every goku ws command

goku ws is short for goku workspace. NAME is a workspace's id or name.

Actions wait and show each step; --no-wait returns as soon as Goku has started one, and goku ws show NAME follows it. ls and show take --json. The command exits 1 when a step failed, 2 for an unknown service, template or color, and 3 when the console can't be reached. It talks to the console at http://localhost:8080; GOKU_CONSOLE_URL points it elsewhere.

Workspaces from your AI assistant

Goku's MCP server has five workspace tools, so Claude Code, Cursor, VS Code and other MCP clients can drive the same workspaces:

Ask, for example:

• Save Glue, S3 and Athena as a workspace called Data day.
• Pause my Serverless day workspace and switch to Data day.
• Which workspace is using the most memory right now?

More Goku guides