Getting Started
Open the Forge chat, send your first request, and understand what happens while Forge works.
Forge lives inside your application, so start by opening the app you want to work on in Studio.
Open the chat
Click the chat icon in the application header. A resizable Axellero Forge sidebar opens on the right — drag its edge to make it wider or narrower. Studio remembers whether the panel was open and how wide it was.
Forge is in beta. If you don't see the chat icon, the feature may not be enabled for your environment yet.
Send your first request
Type into the input field ("Ask Forge anything...") and press Enter. Shift+Enter adds a line break.
Describe the outcome, not the steps. Forge will explore your application on its own and ask when something is genuinely ambiguous:
You: I need a support agent for our tour agency. It should answer
questions about bookings using the orders table.
Forge: I found the orders table with 8 fields. Before I build the
agent — should it only read bookings, or also modify them?A few starting points, depending on what you're building:
| You want to... | Try asking |
|---|---|
| Automate a process | "Build a workflow: when an order comes in, check stock and save it to the database" |
| Create an AI agent | "Create an agent that answers questions about our products" |
| Set up data storage | "Create a customers table with name, email, and phone" |
| Connect a service | "Connect our internal REST API at api.example.com" |
| Find a problem | "Why did the last workflow run fail?" |
What you'll see while Forge works
Forge shows its work as it goes:
- Status line — what's happening right now: Thinking..., running an action, or generating the reply.
- Work plan — for larger tasks, a collapsible progress panel above the chat lists the steps Forge has planned and checks them off as it goes.
- Actions log — each reply notes the actions Forge took (Used N tools). Expand any entry to see exactly what was read or changed.
- Changes made — after Forge modifies something, a summary lists every created or updated item, grouped by area (Workflow, Database, Configuration...). Each item is a link — click it to jump straight to that workflow, table, or setting in Studio.
Before any of those changes happen, Forge asks for your approval — see Approvals.
Need to interrupt? While Forge is working, the send button becomes Stop — click it to halt the current run. And if you close or reload the page mid-task, the run keeps going: reopen the chat and you'll see it still in progress.
Conversations
Each chat is a conversation scoped to the current application. Click Conversations in the sidebar header to see your history or start a New chat; Forge reopens your most recent conversation automatically.
Starting a fresh conversation is a good habit when you switch to an unrelated task — Forge won't drag context from the previous topic into the new one.
Chat settings
Below the chat you'll find three settings, each remembered for you:
- Model — by default Forge uses the model configured for your platform. You can plug in your own OpenAI-compatible endpoint instead; the API key stays in your browser and is never stored on the server.
- Audience — how Forge talks to you: For beginners (plain language), For developers (default), or For experts (terse, full jargon). This changes only the explanations, not what Forge builds.
- Approval mode — how often Forge asks for confirmation. Covered in Approvals.