Use dypt from your terminal, scripts, and agent workflows
The dypt CLI lets you work with your tasks without opening the web app. It is useful when you are already in a terminal, when you want to inspect a task tree quickly, or when you want an AI coding agent to work from the same task list you use in dypt.
Install the CLI
Install the published npm package globally:
npm install -g @dypt/cli
Check the installed version:
dypt --version
You can also run a one-off command without installing globally:
npm exec @dypt/cli@latest -- --version
To update an existing global install, run the install command again:
npm install -g @dypt/cli@latest
Sign in
Start the browser-based sign-in flow:
dypt auth login
The CLI shows a short approval URL. Open it in your browser, sign in to dypt if needed, and approve the device. Once approved, the CLI stores a local token for future commands.
Check which account the CLI is using:
dypt auth status
Sign out when you no longer want this device to have access:
dypt auth logout
You can also manage CLI access from dypt itself. Open the CLI settings page in your account to review approved devices and revoke tokens if a device is lost, shared, or no longer trusted.
Read tasks
List recent tasks:
dypt task list
Filter task lists when you want a narrower working set:
dypt task list --status "in progress" dypt task list --priority high dypt task list --archived all dypt task list --title-contains "launch"
You can also filter by hierarchy. Use --parent for direct children of a task, or --scope for every descendant under a branch:
dypt task list --parent Work dypt task list --scope "CLI" --show-path
For planning views, combine filters in a single command:
dypt task list --scope "CLI" --status "not started" --priority high --show-path dypt task list --deadline-from 2026-04-24 --deadline-to 2026-05-01 dypt task list --blocked only
Search for a task:
dypt task search "launch checklist"
Show one task and its immediate context:
dypt task get 12345
Show a task tree:
dypt task tree 12345
Summarise a branch, including child counts, status breakdowns, due date signals, and branch health signals:
dypt task summary 12345
Create and update tasks
Create a task:
dypt task create "Record CLI walkthrough"
Create a child task under an existing parent:
dypt task create "Draft script" --parent 12345
Set planning metadata as you create the task:
dypt task create "Record CLI walkthrough" --priority high --deadline 2026-05-01 --expected-time 45
Update task metadata:
dypt task update 12345 --status "in progress" dypt task update 12345 --priority high dypt task update 12345 --status completed
Update multiple tasks at once:
dypt task update-many 12345 12346 12347 --status completed
Work with notes
Set the note for a task:
dypt note set 12345 "## Context This task is ready for implementation."
Read a task note:
dypt note get 12345
Search notes:
dypt note search "implementation detail"
Plan work from the terminal
Planning commands help you decide what to do next without manually reading a large task tree.
Find likely next actions under a branch:
dypt task plan next-actions 12345
Get focus signals for a branch:
dypt task plan focus 12345
These commands are designed to work well for both people and AI agents. They surface task status, blockers, missing planning metadata, priorities, deadlines, and branch focus information in a concise format.
Work with AI agents
The CLI is especially useful when you use tools such as Codex, Claude Code, Cursor, ChatGPT, Gemini, or other coding agents. Instead of copying task details into a prompt, you can ask the agent to use the CLI directly.
For example, you can tell an agent:
Use the dypt CLI as the source of truth. Start by running: dypt task get 12345 dypt task tree 12345 dypt task plan next-actions 12345 Update the task notes with any durable implementation context, and mark the task complete when the work is finished.
This lets agents:
- read the current task brief
- inspect subtasks and blockers
- create implementation subtasks
- update notes with durable handoff context
- mark work in progress or complete
- keep dypt in sync with the code changes they make
Clean up task trees
Cleanup workflows help you make larger structural changes to a task tree without manually moving and archiving each task one by one. They are designed for repeatable cleanup patterns, such as collapsing an old wrapper task after moving its surviving children into the right home.
Named cleanup workflows preview by default, so you can review the generated changes before applying them.
For example, the relocate-and-collapse workflow moves the surviving direct children out of a wrapper task and archives the wrapper only when that leaves it empty:
dypt task cleanup relocate-and-collapse --wrapper "Old wrapper" --tasks 42 43 --to "Canonical branch"
Apply the same workflow only when the preview matches your intent:
dypt task cleanup relocate-and-collapse --wrapper "Old wrapper" --tasks 42 43 --to "Canonical branch" --yes
The lower-level cleanup preview and cleanup apply commands are for explicit cleanup plan JSON, not task refs. Most users and agents should prefer the named workflow commands.
dypt task cleanup preview --plan '{"operations":[{"type":"archive","taskIds":[42]}]}'
Need help?
For the public CLI landing page, examples, and npm package link, visit dypt.app/cli.
If you get stuck, contact us from the dypt site or check that you are using the latest CLI version:
dypt --version