Aura-IDE is an open-source desktop AI programming application, officially positioned as an "AI coding harness", that is, a framework environment for doing code work around AI. Its core philosophy is "Code with AI. Work your way.", letting developers collaborate with AI to complete programming tasks in their own way.
Main capabilities:
- Project and repository awareness: Aura can read and search your code repository, track references, and use the project's own tools.
- Precise editing and safety approval: It can perform inspectable "surgical" edits; when human approval is enabled, you can view the proposed diff before writing. It also provides additional control methods such as Read Only and Plan review.
- Verification and recovery: During the work process, it displays tool activity, terminal output, check results, and a final report (including failure cases).
- Agent work isolation: The work of writable Agents is carried out in an isolated worktree, and retained changes need to be reviewed through Aura before being applied.
- Workspace control: Workspace panels can be hidden and reopened while retaining files and conversations.
- Extended tools: MCP servers can be connected within the same application, and Git and terminal tools can be used.
How to use:
- Windows: Download and run the latest .exe from Releases. The installer has built-in Python, installs per user, and can be updated within Aura.
- macOS and Linux: Run from source using Python 3.10+, see the installation guide. The basic process is: open a project folder; connect a model (add a hosted API key in Settings → API Keys, or connect a running local server in Settings → Models); make a small change request, for example "find the function that needs testing, add a test, run it, and explain the results"; then review the changes and view the reported verification results.
Agents and Teams: After enabling Agents, Aura can form teams in a conversation or use a saved Team. An Agent is a reusable assistant; a Team is a savable workflow that connects multiple Agents and assigns tasks to them. An Agent can inherit the current model, or use other configured hosted or local models. Through conversation, you can have Aura create reusable processes, for example "create a reusable workflow to implement the requested changes, test it, and then review it". Creating a workflow only saves the process without executing it; you can inspect the graph cards, request modifications (such as "add a security review") or undo edits, then click Run and provide a task to use it. The Agents & Teams library supports search, and you can open a Team to adjust the graph structure, edit task assignments, and add branches and assistants; saved workflows and their chat cards remain available after restart.
Model selection: Local coding requires the model/server to support tool-calling and have sufficient context. Hosted and local models can be mixed within the same Team.
Mobile: Aura Companion allows you to browse projects, send messages, and follow up on work on your phone, and needs to be paired with the desktop via a local or hosted relay. The desktop is responsible for continuously running the project, so it must remain powered on.
License and fees: Aura is free under the MIT license. Hosted providers charge separately for API usage. Local models require running a server such as Ollama, LM Studio, or llama.cpp, and Aura will connect to it and automatically discover its models.
Other information:
- Official entry points are provided for Windows downloads, tours, documentation, and the Discord community.
- You can support project development through GitHub Sponsors or Buy Me a Coffee.
- The project uses Aura itself for development.
- More facts such as specific pricing and the complete platform support matrix are not yet provided in the materials.
