Skip to main content
Zenflow organizes AI-driven work into a small set of primitives. Understanding how they fit together makes everything else click.

Repositories

A repository is a connected codebase. Zenflow clones it, tracks its branches, and uses it as the scope for every Code task.

Deep dive

Repository management, settings, and team configuration

Tasks

A task is the fundamental unit of work. When you create a task, Zenflow spins up an isolated environment and assigns an agent to execute it.
PropertyDescription
ModeCode (git repo + worktree) or Work (local folder, no repo)
WorkflowDetermines how the agent approaches the work
BranchDedicated branch created automatically
WorktreeIsolated directory at .zenflow/worktrees/{task_id}
StateTo Do → In Progress → In Review → Done (or Cancelled)
Tasks support parallel execution — you can run as many as your machine allows, each in its own worktree.

Deep dive

Task creation, task view, states, artifacts, and actions

Workflows

Workflows define the execution strategy for a task — how much planning happens, whether specs are generated, and how verification works.
WorkflowWhen to use
AutoMost tasks — agent decides the approach
Fix a BugRegressions and production issues
Spec FirstReview the plan before coding starts
Requirements FirstLarge features needing PRD → spec → implementation
Multi-modelDifferent models for planning, building, and review
CustomYour own Markdown-defined workflow templates
Work mode has its own set: Auto, Brainstorm, Deep Brainstorm, Research, and Write.

Deep dive

Workflow comparison, when to use what, and creating tasks

Git Worktrees

Every Code task runs in its own Git worktree — a separate checkout of the repository with its own branch. This is how Zenflow achieves true parallel execution without conflicts.
  • Created automatically at .zenflow/worktrees/{task_id}
  • Agents read, write, and commit inside the worktree
  • Your main working directory is never touched
  • Cleaned up when you archive or delete a task

Deep dive

Worktree lifecycle, IDE integration, and best practices

Agents and Presets

An agent is the AI runtime that executes task steps. Zenflow supports multiple agent backends, and you configure them via presets.
ConceptDescription
Agent presetNamed configuration: agent CLI + model + permission mode
Default agentThe preset used for new tasks (configurable in Settings → Default agents)
Step-level overrideAssign a different preset to individual workflow steps
Presets are managed in Settings → Saved presets and can be selected per-task or per-step.

Deep dive

Supported agent CLIs and model options

Verification

Zenflow can run verification scripts after each agent turn to catch issues early:
  • Configured per-repository in Repository Configuration
  • Runs linting, type checks, tests, or custom scripts
  • Failures are fed back to the agent for self-correction
  • Keeps the feedback loop tight so errors don’t compound

Automations

Automations let you schedule recurring tasks — either at the workspace level (cron-style) or inside an existing task (interval-based agent sessions).

Deep dive

Creating and managing scheduled automations

Next Steps

Quickstart

Create your first task in under 5 minutes

Workflows

Choose the right workflow for your task

Zenflow Work

Non-code tasks, automations, and integrations

Repository Configuration

Verification scripts, setup commands, and presets