> For the complete documentation index, see [llms.txt](https://docs.locusai.dev/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.locusai.dev/core-strengths/unified-interface.md).

# Unified Interface Across AI Clients

One Locus command interface across Claude and Codex -- switch providers without changing workflows.

Locus gives you one CLI to plan, execute, review, and iterate -- regardless of whether you use Claude or Codex under the hood.

## What "One Interface" Means

The workflow surface stays the same across providers:

* Plan: `locus plan`
* Execute: `locus run`
* Review: `locus review`
* Iterate: `locus iterate`
* Observe: `locus status`, `locus logs`

Provider choice is a configuration toggle:

```bash
# Use Claude
locus config set ai.model claude-sonnet-4-6

# Switch to Codex
locus config set ai.model gpt-5.3-codex
```

When `ai.model` changes, Locus infers the provider automatically. All commands continue to work identically.

***

## Side-by-Side Workflow Equivalence

The command surface is intentionally identical across clients:

| Workflow Stage         | Claude                                                 | Codex                                     |
| ---------------------- | ------------------------------------------------------ | ----------------------------------------- |
| Select model           | `locus config set ai.model claude-sonnet-4-6`          | `locus config set ai.model gpt-5.3-codex` |
| Plan work              | `locus plan "Add webhook retries" --sprint "Sprint 7"` | Same command                              |
| Execute sprint         | `locus sprint active "Sprint 7"` then `locus run`      | Same commands                             |
| Review PRs             | `locus review`                                         | Same command                              |
| Iterate on feedback    | `locus iterate --sprint`                               | Same command                              |
| Resume interrupted run | `locus run --resume`                                   | Same command                              |

The only required change is model selection. Everything else stays unchanged.

***

## Supported Models

### Claude (Anthropic)

| Model             | Alias                                  |
| ----------------- | -------------------------------------- |
| Claude Opus 4.6   | `claude-opus-4-6` or `opus`            |
| Claude Sonnet 4.6 | `claude-sonnet-4-6` or `sonnet`        |
| Claude Haiku 4.5  | `claude-haiku-4-5-20251001` or `haiku` |

### Codex (OpenAI)

| Model               | Alias                 |
| ------------------- | --------------------- |
| GPT-5.4             | `gpt-5.4`             |
| GPT-5.4 Pro         | `gpt-5.4-pro`         |
| GPT-5.3 Codex       | `gpt-5.3-codex`       |
| GPT-5.3 Codex Spark | `gpt-5.3-codex-spark` |
| GPT-5.2 Codex       | `gpt-5.2-codex`       |
| GPT-5.1 Codex Max   | `gpt-5.1-codex-max`   |
| GPT-5.1 Codex       | `gpt-5.1-codex`       |
| GPT-5.1 Codex Mini  | `gpt-5.1-codex-mini`  |
| Codex Mini Latest   | `codex-mini-latest`   |

***

## Automatic Provider Inference

Locus infers the provider from the model name:

1. Known Claude aliases (`opus`, `sonnet`, `haiku`) → Claude
2. Known Codex models (`gpt-5.4`, `gpt-5.4-pro`, `gpt-5.3-codex`, etc.) → Codex
3. Model name starts with `claude-` → Claude
4. Model name contains `codex` → Codex

You rarely need to set `ai.provider` directly -- setting `ai.model` is sufficient.

***

## Per-Command Override

Override the model for a single run without changing your configuration:

```bash
locus run --model opus
locus run 42 --model gpt-5.3-codex
```

The `--model` flag takes precedence for that execution only.

***

## Environment Variables

AI provider authentication is handled through environment variables or provider CLI login, not through Locus:

```bash
# Claude
export ANTHROPIC_API_KEY="sk-ant-..."
# Or use: claude auth login

# Codex
export OPENAI_API_KEY="sk-..."
```

Locus does not store provider API keys.

***

## Recommended Adoption Path

1. Start with [Quickstart](/getting-started/quickstart.md).
2. Standardize team docs around `plan -> run -> review -> iterate`.
3. Treat model/provider as a configuration toggle (`ai.model`), not a workflow fork.

## Related Docs

* [How Locus Works](/workflow-deep-dives/how-it-works.md)
* [Built-In Tools](/core-strengths/overview.md)
* [Auto-Approval Mode](/core-strengths/auto-approval-mode.md)