Plans

Build, save, and run multi-step workflows that orchestrate agents, tools, and skills.

Overview

Plans are reusable workflow templates. Each plan defines a sequence of steps, which agents execute them, which tools they use, and how data flows between steps. Plans can be created through the chat interface or built manually.

Creating Plans

From Chat

The easiest way to create a plan is through the chat panel. Describe what you want done in plain language, and Agentican will generate a plan with steps, agent assignments, and tool selections. You can refine the plan through conversation before running or saving it.

Saving Plans

After generating a plan from chat, you can save it for reuse. Saved plans appear on the Plans page where you can edit, run, or delete them. Each save creates a new version, so you can iterate on plans without losing previous configurations.

Plan Structure

Every plan consists of:

• Name and description — Identifies the plan
• Steps — Individual units of work, each assigned to an agent
• Parameters — Input variables that can be customized at run time
• Schedule (optional) — Automate execution on a recurring basis

Steps

Each step in a plan has:

• Instructions — What the agent should do
• Agent — Which agent executes this step
• Tools — External integrations available to the agent
• Skills — Additional expertise for the agent
• Dependencies — Which steps must complete first
• Approval required — Whether the output needs human review

Dependencies & Execution Order

Steps define dependencies on other steps. Steps without dependencies (or whose dependencies are all complete) can run in parallel. This creates a directed acyclic graph (DAG) that determines execution order.

Control Flow

In addition to regular agent steps, plans support two control flow step types:

Loop Steps

Loop steps repeat a set of body steps for each item in a collection. When a loop runs, it creates a separate copy of its body steps for each iteration. Each iteration is tracked independently in the task view with its own status, tokens, and output.

For example, a loop step might iterate over a list of customers, running "Draft onboarding email" for each one. The loop item is passed to the body step's instructions so the agent has context for each iteration.

Branch Steps

Branch steps enable conditional execution based on previous step outputs. Only the matching branch path runs — other paths are skipped. This lets you build plans that adapt to different scenarios without running unnecessary steps.

Versioning

Every time you save a plan, Agentican creates a new version. This means you can iterate on a plan — adding steps, changing agents, adjusting instructions — without losing previous configurations. Tasks created from a plan reference the specific version that was current when they ran.

Parameters

Plans can accept input parameters that customize behavior at run time. Parameters are referenced in step instructions using {{parameter_name}} syntax and resolved before execution begins.

Each parameter has:

• Name — The variable name used in templates
• Label — A human-readable label shown in the run dialog
• Type — Text (free-form input) or select (choose from predefined options)
• Default value — Pre-filled when running the plan
• Required — Whether the parameter must be provided

For example, a "Competitor Analysis" plan might accept a competitor_name parameter so it can be reused for different companies. When running the plan, a dialog prompts for parameter values.

Scheduling

Saved plans can be configured to run on a schedule. Set a frequency — daily, weekly, or monthly — along with a specific day, time, and timezone. Agentican will automatically execute the plan at the specified times. Scheduled runs appear as tasks just like manual runs.

Managing Plans

The Plans page shows all your saved plans with:

• Plan name and description
• Step count
• Schedule status (if configured)
• Actions: Run, Edit, Delete

Deleting a plan is a soft delete (archive) — existing tasks created from the plan are not affected.