AgentsInFlow
AgentsInFlow

Execution & Config

Configure engines, branching, and execution control behavior.

Updates

Auto-update checks are enabled in packaged builds.

  • First background check runs about 15 seconds after app launch.
  • Ongoing background checks run every 4 hours.
  • Manual check is available from AgentsInFlow → Check for Updates... on macOS.

Git Branching

When starting an execution, you can enable branching to run the task in an isolated git worktree. This keeps your main branch clean while the agent works.

How to use

1

Enable "Use Branch" toggle in the run dialog

2

Select source branch (defaults to current HEAD)

3

Optionally specify a target branch name (auto-generated if empty)

Git worktrees share the same git objects as your main repository. They're fast to create and don't take much disk space. Changes remain isolated until you manually merge them.

Engine Configuration

Each project can have per-engine configuration. Settings are written to your project directory:

Claude .claude/settings.local.json
Codex .codex/config.toml
Cursor .cursor/settings.json

Configuration Options

Model selection
Permission mode
Sandbox settings
MCP servers (Claude)

Model Selection

Choose which model to use per engine. The model dropdown in the run dialog shows available models for the selected engine. Set a default model in project settings and override it at execution time.


3-Level Settings Cascade

Engine settings follow a three-level hierarchy where more specific levels override less specific ones:

1

Global

Lowest priority

Default for all projects. Set in app-level Settings.

2

Project

Per-project, per-engine overrides.

3

Execution

Highest priority

Per-run overrides chosen in the run dialog.


Auto-Allow Tools

Configure which tools are automatically allowed for each project. When enabled, the agent skips permission prompts for these tools, reducing interruptions during execution. Tool permissions are written to the engine config file and apply to all executions in that project.