Agent Presets

What Are Agent Presets?

Agent presets are named model configurations stored in Settings → Saved presets. Each preset defines:

Coding agent — which runtime to use (Zencoder, Claude Code, Codex, Gemini)
Model — which model the agent uses (e.g., Claude Opus, Gemini Flash, Claude Sonnet)
Permission Mode — execution mode and tool access level

Zenflow saved presets dialog showing preset name, coding agent, model, and permission mode fields

You can create multiple presets — for example, opus-planner, flash-builder, sonnet-reviewer — and assign them to different workflow steps.

Assigning Presets to Workflow Steps

In a custom workflow file (.zenflow/workflows/), use the  comment to bind a preset to a specific step:

# Multi-Agent Feature Workflow

## Configuration
- **Artifacts Path**: {@artifacts_path} → `.zenflow/worktrees/{task_id}`

---

## Workflow Steps

### [ ] Step: Planning
<!-- agent: opus-planner -->
Analyze the task requirements. Produce a technical specification in `{@artifacts_path}/spec.md` covering:
- Architecture decisions and trade-offs
- Affected files and modules
- Edge cases and error handling
- Verification criteria

### [ ] Step: Implementation
<!-- agent: flash-builder -->
Implement the changes described in `{@artifacts_path}/spec.md`. Follow the project's existing patterns and conventions.
- Write code that satisfies the spec
- Include unit tests for new logic
- Update `{@artifacts_path}/plan.md` with completed items

### [ ] Step: Review
<!-- agent: sonnet-reviewer -->
Review all changes from the Implementation step against `{@artifacts_path}/spec.md`.
- Check for correctness, edge cases, and regression risks
- Verify scope discipline — no unnecessary changes
- Confirm tests cover the new behavior
- Record findings in `{@artifacts_path}/review.md`

When the task runs, Zenflow automatically switches to the assigned preset at each step boundary.

Sequential Multi-Model Review

For higher-stakes changes, add multiple review steps — each runs after the previous one, so each reviewer sees the last reviewer’s output:

### [ ] Step: Review — Architecture (Claude Opus)
<!-- agent: opus-reviewer -->
Deep review of architectural decisions, contract adherence, and regression risks.
Record findings in `{@artifacts_path}/review-architecture.md`.

### [ ] Step: Review — Implementation (Claude Sonnet)
<!-- agent: sonnet-reviewer -->
Implementation-level review: code quality, test coverage, scope discipline.
Record findings in `{@artifacts_path}/review-implementation.md`.

Default Agents

Configure the default agent preset for all new tasks in Settings → Default agents:

Setting	Description
Default agent & config	The preset used for new task implementation
Reviewer agent & config	The preset used for review steps
Code task instructions	Instructions injected into every code task prompt

When to Use Presets

Scenario	Fit
Sequential workflow steps with different models	Best fit
Recurring workflows reused across tasks	Best fit
Need parallel execution within a phase	Use Subagent Pipelines instead
Want model assignments in the workflow file itself	Use Subagent Pipelines instead

Get Started

Zenflow

Agents

Workflows

Customizing

Models

Integrations & MCP

IDE Plugin

CLIs

Admin

Resources

What Are Agent Presets?

Assigning Presets to Workflow Steps

Sequential Multi-Model Review

Default Agents

When to Use Presets

​What Are Agent Presets?

​Assigning Presets to Workflow Steps

​Sequential Multi-Model Review

​Default Agents

​When to Use Presets

What Are Agent Presets?

Assigning Presets to Workflow Steps

Sequential Multi-Model Review

Default Agents

When to Use Presets