AgentsInFlow
AgentsInFlow
Download

Engine & Model Configuration

Configure AI engines with a 3-level settings cascade, model selection, and auto-allow tools.

3-Level Settings Cascade

Engine settings follow a three-level hierarchy. More specific levels override less specific ones:

1

Global Settings

Lowest priority

Default configuration for all projects and all engines. Set in the app-level Settings panel.

2

Project Settings

Per-project, per-engine configuration. Overrides global settings for that project. Set in the project settings panel.

3

Execution Settings

Highest priority

Per-execution overrides chosen in the run dialog. Model, branch, and permission mode can all be changed at run time.

Engine Configuration

Each engine (Claude Code, Codex, Cursor) can be configured independently. Settings are written to engine-specific config files in your project directory:

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

Model Selection

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

Auto-Allow Project Tools

Select which tools are automatically allowed in Claude project settings. When enabled, the agent does not need to ask for permission before using these tools, reducing interruptions during execution.

  • Configure allowed tools per project in project settings
  • Tool permissions are written to the engine config file
  • Applies to all executions in that project

Compact Tool Responses

Toggle compact mode to reduce output noise from tool calls. When enabled, tool responses are collapsed by default in the terminal, keeping the output cleaner and easier to scan. This is the default for ticket-based executions.