OMADS

New Project

Project name

Project directory

Settings

Primary builder

All normal chat tasks go to this builder until you change the setting again.

Project directory

Automatic post-change review

When enabled, OMADS runs an automatic breaker step after builder-created code changes and returns findings to the builder for a follow-up fix decision. Today this uses Codex after Claude builds and Claude Review after Codex builds.

Project tools

Open the current Git working tree diff for this project.

Inspect API

FastAPI documentation is available locally while OMADS is running.

Swagger UI ReDoc OpenAPI JSON

Theme

Color theme

Choose the interface theme that feels best for long coding sessions.

LAN access

Open for local network

When enabled, OMADS binds to all network interfaces so you can open it on your phone. Requires a server restart to take effect.

Manual Review Pipeline

Reviewer 1

The first reviewer inspects the selected scope and also performs the final synthesis in step 3.

Reviewer 2

Reviewer 2 must be the other agent so the manual review stays cross-checked.

Step 3

The final synthesis and optional fix plan always go back to Reviewer 1.

The Review button is a separate manual feature. It does not replace the normal builder → breaker loop used during coding tasks.

Model

Claude model

Permissions

Permission mode

This setting is passed directly to Claude Code via `--permission-mode`, so the selected mode now takes effect on the next run.

Reasoning

Effort

Controls how deeply Claude reasons. Higher values usually improve results but take longer.

Codex role in OMADS

Codex CLI Builder and reviewer, depending on the active flow

Codex can act as the selected primary builder, as the automatic breaker after Claude-built code changes, and as either manual reviewer depending on the review settings.

Model

Codex model

Select Model

Access legacy models by running codex -m <model_name> or in your config.toml.

1. gpt-5.4 default Latest frontier agentic coding model. 2. gpt-5.4-mini Smaller frontier agentic coding model. 3. gpt-5.3-codex Frontier Codex-optimized agentic coding model. 4. gpt-5.2-codex Frontier agentic coding model. 5. gpt-5.2 Optimized for professional work and long-running agents. 6. gpt-5.1-codex-max Codex-optimized model for deep and fast reasoning. 7. gpt-5.1-codex-mini Optimized for codex. Cheaper, faster, but less capable.

Pick one of the same models shown by the Codex CLI model picker. The selected entry is applied to every Codex run in OMADS.

Reasoning

Reasoning effort

Select Reasoning Effort

Matches the Codex CLI effort selector and controls how deeply Codex thinks before responding.

1. low Fastest, least thorough. 2. medium Balanced default for most tasks. 3. high default Strong default for most tasks. 4. xhigh Most thorough, slower.

Higher values usually improve results but take longer.

Execution

Execution preset

This preset controls Codex sandboxing and approval behavior for every Codex run in OMADS.

Fast Mode

Fast mode favors lower-latency responses.

The selected primary builder handles normal chat tasks and coding work.
Automatic post-change review currently runs after builder-created code changes and returns findings to the active builder for a follow-up fix decision.
Today this uses Codex after Claude-built changes and Claude Review after Codex-built changes.
The Review button runs the manual three-step pipeline configured in the Review settings tab.
This tab shows the Codex-specific settings used in the current OMADS flow.

Repository Diff

Loading current working tree status...

Loading diff...

New Project

Start code review

Settings

Repository Diff

Open on your phone

GitHub

Git