Open a folder and give the agent a finish line
- Install the desktop app for your operating system and sign in to LoopCode.
- Choose the project folder containing your source code. Review your working changes before starting a new task.
- Describe the expected behavior, relevant files if you know them, and a check that demonstrates success.
- Use Agent mode to begin work, or Plan mode when you want an approach before implementation. Choose permissions suited to the project.
- Inspect the diff and command output. Run the affected workflow and confirm anything the agent could not verify.
Trace how the order API validates quantities.
Fix the case where a negative quantity reaches storage.
Preserve the current response format. Add a regression test.
Run the existing API tests and show the final diff.
Do not deploy or change production data.Follow the work from context to verification
| Part of the workflow | What you use it for | What it does not establish on its own |
|---|---|---|
| Project conversation and file context | Explain a module, scope a feature, or identify the cause of a bug | That an assumption matches every caller or deployment |
| Changes and command output | Inspect the patch and the tools run against your project | That every edited behavior has been tested |
| LSP code intelligence | Find definitions, references, and diagnostics from a configured language server | A guarantee of runtime correctness or support for an unconfigured language |
| Desktop browser | Inspect and interact with an accessible running web app | Backend correctness, a complete accessibility audit, or production readiness |
Build more than a landing page
Use the workspace to investigate application code, backend services, internal tools, or business automations. The useful unit of work is a change with observable behavior: an API validation rule, a data import with fixtures, a dashboard filter, or a failing test with a clear expected result.
The agent uses the project’s tools and dependencies. A database, external integration, or test environment still needs to be configured. Keep production credentials and destructive operations out of an exploratory task; review the patch as you would another contributor’s work.
Local execution and hosted models have different boundaries
The desktop engine operates against your selected project environment. Your repository and conversation storage are local, while relevant prompts and code context are sent to the selected hosted model to generate a response. Hosted inference requires connectivity.
A local desktop agent is therefore different from a fully offline coding model. Before opening a sensitive repository, confirm that sharing its relevant context with the model service is permitted. Tool permissions control actions; they are not a claim that all context stays on your computer.
Choose the interface that fits your working day
Desktop installers are available for macOS 13 or newer on Apple silicon and Intel, Windows 10 or newer on x64, and Linux x64. The download page and installation guide contain the current release assets, first-launch notes, and installer requirements.
Use the desktop app for a graphical workspace and browser checks, the terminal for interactive or bounded headless tasks, or VS Code for a conversation beside your code. Sharing an account and plan does not mean every interface has identical tools; browser verification requires the desktop integration.
Start with a task you can verify
Free starter usage lets you try the workflow. For more capacity, compare the current Pro and Max limits and prepaid pricing. Model availability can change; use the account’s available models rather than a static model list as the source of truth.
Keep a trial small enough to review. Record the initial repository state, the final patch, the checks run, and any remaining work. Those details tell you more about fit than a generated screenshot or an unqualified claim of being the best coding agent.
KEEP GOING
A useful next step.
Put it to work in your project.
Desktop, CLI, and VS Code. One LoopCode account. Free to start.