Devora

English | 简体中文

Devora is a project-aware workflow for AI coding agents. It adds a consistent engineering process to an existing repository, helping agents plan changes, implement them, verify the result, and prepare a clear delivery record.

Devora works with the coding agent you already use. It installs native commands and project-level workflow files without introducing a separate model runtime.

Why Devora

• Consistent delivery — guide each change through Design, Build, and Review.
• Project awareness — work from the repository's code, documentation, and engineering conventions.
• Role-specific guidance — support frontend, backend, and full-stack development.
• Traceable changes — keep plans, tasks, progress, review results, and handoff information together.
• Team customization — apply reusable rules and skills at company, project, and personal scopes.
• Agent portability — use the same workflow across supported coding-agent integrations.

Getting started

1. Install Devora

Devora requires Python 3.11 or later. Install the CLI with uv:

uv tool install devora-cli

Confirm the installation:

devora --version

2. Initialize a project

Run devora init from the repository you want to use:

cd /path/to/project
devora init . --integration codex --role frontend --language en

Choose the integration that matches your coding agent and the role that matches the repository. Initialization adds Devora's workflow files; it does not modify application code.

3. Start your first change

Open the project with your coding agent and invoke Devora with a clear objective and any relevant sources:

$devora

Implement order cancellation.

Reference material:
- ./requirements/order-cancel/
- ./docs/order-api.md
- https://www.figma.com/design/...

Devora accepts direct descriptions as well as references to project files, directories, specifications, API documents, issue links, and UI designs.

For integrations that use slash commands, invoke /devora instead of $devora.

4. Review the delivery record

Workflow artifacts are stored in a named change directory:

.devora/changes/order-cancel/
├── design.md
├── tasks.md
├── summary.md
├── review.md
└── handoff/

Use summary.md to follow progress, review.md for the final review result, and handoff/ for integration and delivery information.

Workflow

Design  →  Build  →  Review  →  Human approval

The main entry point resumes the appropriate stage from the current project state. Advanced users can invoke a stage directly:

Development roles

Select one role for each initialized repository:

Manage the active role with the CLI:

devora role list
devora role current .
devora role use backend .

Customization

Devora loads Markdown rules and agent skills from three scopes:

Each scope can contain:

rules/                  Markdown guidance files
skills/<name>/SKILL.md  Agent skills

Personal configuration can also include preferences.md.

Supported integrations

List, switch, or upgrade integrations:

devora integrations
devora integration use claude .
devora integration upgrade .

Run devora <command> --help for the complete CLI reference.

Development

uv sync --extra test
uv run pytest
uvx ruff format --check src tests scripts/python scripts/ci
uvx ruff check src tests scripts/python scripts/ci

Additional documentation:

• Architecture
• Changelog
• Issue tracker

Project status

Devora is currently in alpha. Interfaces and generated workflow files may evolve before a stable release.

License

Devora is licensed under the Apache License 2.0.