An AI code editor built aroundthe whole development loop.

The editor is built on Monaco, the editing engine behind Visual Studio Code. You get syntax highlighting, multi-cursor editing, bracket matching, folding, and a file tree with create, rename, delete, and drag-and-drop operations. Open files stay in tabs so you can move between related parts of a project without losing your place.

The explorer includes fuzzy search across file names and a recent-work view. A problems panel collects diagnostics from the editor and shows file and line references so you can jump back to the source of an error.

• Monaco editing with multi-tab buffers and configurable editor settings.
• Workspace file tree with validated create, rename, delete, and move operations.
• Fuzzy file-name search, recent files, and a unified problems list.
• Account-scoped persistence, so a signed-in workspace can be reopened later.

The assistant can work across multiple files instead of returning a single pasted snippet. In Agent and Edit modes it can list and read the workspace, search for relevant files, inspect diagnostics, and create, modify, rename, or delete files as part of the task.

File changes are validated against the current workspace before they are written, and the assistant reports which paths it touched. The main assistant flow is live: it edits the workspace rather than stopping at a suggestion. Inline editor actions can present a preview for review in some surfaces.

• Read, search, list, write, create, rename, and delete tools with validation.
• Streaming activity so you can see what the agent is inspecting or changing.
• Diagnostics-aware inspection of the files involved in a task.
• Ask mode for questions; Agent and Edit modes when you want changes.

Azemax has two AI modes. Azemax Default sends requests through the Azemax backend and uses the service’s model routing. Bring Your Own Key sends requests from your browser to the provider you configure, using your own credentials and model selection.

The model list shown in settings is a starter catalog. Providers add, rename, and retire models on their own schedule, so the exact list can change. When a provider exposes a model-discovery endpoint, Azemax can fetch the current catalog; otherwise you can enter a model identifier manually. The Custom provider accepts any HTTPS endpoint that implements an OpenAI-compatible chat completions API.

Bring-your-own-key usage is billed by your provider, under that provider’s terms. Azemax does not add a markup to provider usage and does not receive your provider key when you use BYOK mode.

Step-by-step provider setup, including the exact settings path and custom-endpoint requirements, is in the provider configuration guide.

The Run action sends the active file to the hosted execution service and opens the output panel with stdout, stderr, compile errors, and the exit status. The hosted product uses an isolated Wandbox sandbox rather than executing code on the application server.

Hosted execution supports Python, JavaScript, TypeScript, C, C++, Java, Ruby, PHP, Perl, Rust, Go, Swift, Lua, Bash, C#, Scala, and R. Compiler and runtime versions are supplied by the sandbox service and can change.

The assistant receives context assembled from your project: the active file, relevant open files, the file tree, selected text, and results from workspace tools. That is why it can refer to real paths and propose changes in the right place instead of guessing from a pasted excerpt.

Context is not unlimited. Large repositories and long conversations can exceed a model’s context window, and different providers accept different input sizes. Azemax may ask the agent to search for specific files rather than sending the entire tree. If a request feels like it is missing context, name the file or folder you want the assistant to inspect.

Azemax Studio requires an account; there is no guest workspace. You can sign up with email and password or with Google. Email/password accounts must verify their address before the workspace and product APIs unlock. Google sign-in arrives with a verified email and skips the separate verification step.

When you are signed in, projects, settings, and assistant conversations are associated with your account and can be restored on another device. Access is scoped to the authenticated user; the workspace is not a public sharing surface. Multiplayer editing and public project links are not part of this beta.

Provider keys are sensitive. By default, a BYOK key lives only in memory for the current tab. If you opt into “Remember on this device,” the key is encrypted with AES-GCM using a key derived from a passphrase you choose; only the ciphertext is stored in IndexedDB. The passphrase and derived key are not persisted.

In BYOK mode, AI requests go to your provider. In Azemax Default mode, requests go to the Azemax backend. Code you run goes to the hosted execution sandbox. The workspace syncs through Firebase when you are signed in. None of this is a local-only or offline product.

Read the Privacy Policy for the data categories, providers, retention practices, and deletion process. Security issues can be reported through the contact page.

Setting expectations is part of shipping an honest developer tool. In the public beta:

• It is a browser workspace, not a replacement for a full local IDE or your existing toolchain.
• It does not run a general-purpose hosted shell or provide production CI runners.
• It does not guarantee that AI output is correct, secure, or license-compatible.
• It does not work offline, and it depends on third-party AI and execution providers.
• It does not yet offer self-service account deletion; deletion requests are handled by email.
• It is under active development, so interfaces and provider catalogs can change.