Switch from Cabinet to Task Machine
Cabinet keeps agent work beside files on your own machine. Task Machine gives people and agents shared Tasks, recurring workflows and an Inbox for decisions.
Prefer the side-by-side comparison?Why people switch
-
Hosted plans are waitlisted
Cabinet's self-hosted product is available, but Cloud Pro and Cloud Max remain on waitlists. Moving for managed hosting means leaving an available self-hosted app, not an unavailable product.
-
Check-ins need Cabinet open
Cabinet's heartbeat check-ins pause when the cabinet closes and resume when the app starts again. These scheduled checks are not background daemons.
-
Provider setup stays local
Cabinet's runtime documentation requires at least one installed AI command-line tool. Its listed adapters use tools you have already installed and authenticated.
What maps to what
Every answer links to its source and shows when we last checked it. The oldest check on this page was November 15, 2026.
| Area | In Cabinet | In Task Machine |
|---|---|---|
| Tasks |
In Cabinet
A task is one prompt, one agent run and one output. Its card opens the transcript, prompt and diff of every page it touched. |
In Task Machine
A person or an agent owns each Task. Its plan, discussion, decisions, runs and history stay together when responsibility changes. |
| Recurring work |
In Cabinet
Routines are YAML files in .jobs/ that repeat a prompt with a named agent on a cron schedule. Each run records its model, effort and output paths. |
In Task Machine
Schedule a Task or Workflow in your timezone. One stable root Task keeps the occurrences together, and Workflows add branches, human questions and verifier checks. |
| Approval |
In Cabinet
The Approval Queue collects proposed tasks and schedules. Auto-approval policies can match action kind, proposing agent, target agent and estimated cost. |
In Task Machine
Decisions arrive in your Inbox with context. Set each agent action to Ask first or Automatic, and add an explicit Human approval step before publishing a report. |
| Memory |
In Cabinet
Each agent reads its memory file at the start of a run and writes to it at the end. You can open and correct memory.md yourself. |
In Task Machine
Keep reusable lessons per agent, project, goal, team and workspace. You can read, correct and restore notes, such as an accepted example or a mistake to avoid. |
| Models |
In Cabinet
Set a model on the persona or override it for a task. Configured fallbacks apply when a request hits a budget cap or rate limit. |
In Task Machine
Auto chooses a model for each stage from the task, model capabilities, pricing and your applicable budgets. You can fix a particular model on an agent instead. |
| Knowledge |
In Cabinet
Notes, files and agents share a folder on your computer. Outputs return as pages beside their source material. |
In Task Machine
Keep reports, drafts and uploaded reference files in the Library. Documents retain versions and links to the work that produced them. |
| Pricing |
In Cabinet
Self-hosting is free, and you bring your own AI accounts. The pricing page says AI is a separate bill, with optional Cabinet AI starting at $10 a month. |
In Task Machine
Usage based, from $29 a month before tax. The whole payment becomes usage that your workspace shares, with no base fee. |
A fact we cannot find or cannot check is left out, and prices can change after the date shown. Task Machine publishes this comparison and is an interested party. Cabinet is named only to identify it, and there is no relationship between the two companies. Has something changed? Tell us.
You may stay with Cabinet if
You may stay with Cabinet if your main need is a local knowledge workspace you can edit and back up as files. Its overview describes viewers for PDFs, spreadsheets and embedded apps beside your notes. The persona format keeps agent configuration in editable, git-tracked files, so moving only one recurring job may fit better than replacing your knowledge workspace.
What you give up
Task Machine's docs do not describe Cabinet's embedded HTML app workspace or an importer for its personas, job definitions and approval history. Keep your original folder as an archive. Moving instructions, reviewing old memory, reconnecting services and rebuilding schedules takes time. Uploading reference files to the Library does not recreate the old application.
There is no trial. You leave free self-hosting for a paid usage commitment, even if you choose a Local Worker. You keep 100% of your revenue, Task Machine takes no cut and never custodies your accounts.
How the switch works
- Copy the whole Cabinet folder before changing anything. Use its file structure reference to retain .agents/, .jobs/ and .cabinet-state/. Include the .chat/ conversations, which are git-ignored by default, rather than assuming a git backup includes them. Keep hidden runtime files private.
- Create your workspace and upload the selected job's reference material to the Library. Connect its services, then configure an agent with the relevant persona instructions. Review old memory before turning it into standing guidance.
- Rebuild one .jobs/ routine as a Workflow. Replace old folder paths with accessible inputs and explicit document outputs. Add requirements a verifier can check and a human approval before any sending or publishing.
- Choose Supervised autonomy and test on copied inputs. Review the result before enabling live actions. Then disable the Cabinet routine with enabled: false and turn off any corresponding heartbeats before enabling the replacement schedule. Confirm the timezone and next occurrence.
Common questions
Is Cabinet available, or is it a waitlist?
Its self-hosted product is available. Cabinet Cloud Pro and Cloud Max are on waitlists. The waitlist applies to those hosted plans, not the whole product.
What happens to my files and agent history?
Keep the original folder, including the hidden paths in Cabinet's file structure reference and its .chat/ directory. Upload selected reference files to the Library. Task Machine's docs do not describe converting Cabinet history into native Tasks or decisions.
How do the costs compare?
Cabinet is free to self-host, with your AI accounts billed separately. Task Machine's pricing starts at $29 a month before tax, with the whole payment funding shared usage and no base fee or trial. Compare one representative job, including your machine and model costs, rather than treating the two payments as the same unit.
Can I keep Cabinet for knowledge work?
Yes. Keep the file-based workspace and move only the job you want to review in Task Machine. Copy the inputs that job needs into the Library. This is a manual transfer, not a promise that the two workspaces synchronize.