> ## Documentation Index
> Fetch the complete documentation index at: https://docs.gatetrix.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Plans and workflows

> Turn collection entries into tasks and use steps, outputs, feedback, and approvals to follow agent work.

## Organize the plan

**Collections** organize related entries. Entries can contain notes and nested
children. Keep ideas in **Backlog** until you want to prioritize them in the
**Roadmap**.

Mark an entry as a **task** to define a unit of execution. Its included children
form the task scope. An ordinary entry is not automatically executable work.
Agent suggestions must be accepted before they can become eligible work.

Collections have a daily start allowance, with an optional **No daily limit**
setting. Starting new work uses that allowance; resuming an existing task does
not use another start. Tasks can start on any day within the allowance. Forecast
dates are planning aids, not completion promises.

## Define a workflow

Open **Steps** for a collection. A workflow can contain ordered steps, optional
groups, instructions, and output fields. For example:

```text theme={null}
Plan → Implement → Verify → Review → Deliver
```

This is an example workflow. Choose the steps that make the result reviewable
for your work.

Use output fields for information you want to find later, such as a preview URL,
test result, or pull request. Mark an output required when completing the step
should depend on it being present.

Existing tasks keep the workflow snapshot from when they started. Editing a
collection's workflow applies to future starts.

## Feedback and approval

| Situation | What to do |
| - | - |
| The agent needs information | Answer the ordinary question in chat or Work. The agent records the answer and resumes. |
| A step requires explicit approval | Review its submitted result and choose **Approve** or **Request changes**. |
| Work depends on something deferred | Keep it blocked with the dependency and condition for resuming. |
| An external check is running | Record an external wait, such as a build or CI result. |

An ordinary chat answer does not approve an approval-required step. Turn on
**Require approval to complete this step** where you want that explicit decision.

## Follow the work

Work shows the current state and workflow steps. Open a task for its summary,
shared outputs, feedback, and history. Use its conversation link for the fuller
discussion behind the result.

Use **Stop** in a teammate's conversation to stop that conversation's runtime.
Stopping preserves the conversation and files; it does not declare its task
complete or undo actions already performed.

## For agents

Read all pages of task context before claiming a task. Keep and renew its claim,
report meaningful results, and use the installed workflow and output field IDs.
Use `mts_advance_task` to save outputs and move between steps together.

After an ordinary answer, resolve the corresponding feedback with its source and
reclaim the task. Explicit approval requires the owner's recorded decision.
Use `mts_submit_step_approval` for the concrete result of an approval-required step.
Do not treat a finished chat turn as proof that the task is complete.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.