Build Agents with Weaver

Weaver is your chat-first builder inside Agent Studio. Instead of wiring everything manually, you describe what you want, and Weaver creates it on the Studio Canvas in real time.

You can start with:

• Prompt-to-Agent: describe your goal in plain English.
• URL-to-Agent: paste an API spec or integration URL.
• Image-to-Agent: drop in a workflow diagram from tools like Miro or Lucid.

What Weaver does for you

You focus on your intent. Weaver interprets it, asks clarifying questions, proposes a plan, and builds the workflow directly on the Canvas. You stay in control the whole time.

Why use Weaver?

Weaver reduces the startup friction of building agents. You can:

• Prototype quickly by chatting your idea into a live agent.
• Save time with auto-selected components wired for you.
• Work seamlessly with chat on the left and Canvas on the right.
• Recover safely using undo and rollback when you need to backtrack.
• Keep your workspace clear with auto-prettify for layouts.

If you want deeper control later, you can always switch back to manual edits in the Canvas. Weaver gets you started fast, Canvas helps you fine-tune.

Key concepts

Knowing the basics will make your Weaver experience smoother:

• An Agent is a runnable automation built in Studio. Learn more in the Agent overview.
• A Workflow is the connected sequence of steps that define what your agent does. See Building workflows.
• Components are the building blocks—API calls, generators, or data transforms. Browse the Component Library.
• Linting means Weaver checks your prompts and data flow for issues, suggesting fixes before errors block you.

Quickstart: build your first agent

Want a complete walkthrough? Follow the Weaver Quickstart guide.

Here’s the high-level flow:

1. Open Weaver in Studio
   Select an agent and click the Weaver tab in the sidebar.

2. Pick your input

   • Type a goal (Prompt-to-Agent)
   • Paste an API spec (URL-to-Agent)
   • Upload a diagram (Image-to-Agent)

3. Answer clarifying questions
   Weaver may ask about data, audience, or steps you need.

4. Review the plan
   You can edit, refine, or reject suggested steps.

5. Watch the Canvas build
   Components appear and connect automatically as you chat.

6. Test in Studio
   Run the agent, review inputs/outputs, and refine.

7. Deploy to Runtime
   When it works, deploy your agent through Runtime as an API or web agent.

Skip ahead with templates

Start with a prebuilt template. Open it in Studio, then chat with Weaver to adapt it to your needs.

Core features

Feature	What you get
Multiple input types	Start from a prompt, API spec, or diagram
Live Canvas updates	See workflows build visually as you chat
Linter & auto-fixes	Catch unclear prompts, data mismatches, and risky steps
Undo & rollback	Safely revert changes or step back in history
Prettify layout	One-click layout cleanup for clarity
Fix with AI	Repair broken or misconfigured components instantly
Zero-code deployment	Publish directly to Runtime

Debugging with Weaver

When things don’t go as planned:

• Click any red-highlighted block to inspect details.
• Use Fix with AI in chat to repair configuration.
• Multi-select blocks with Cmd/Ctrl + click to fix together.
• Roll back to an earlier state if you want to undo.

For a structured approach, see Debugging Agents.

Troubleshooting tips

• Workflow not matching your goal → Be explicit in your prompt or diagram. Add constraints or expected steps.
• Invalid parameter suggested → Cross-check against the Component Library.
• Agent run fails → Test blocks individually, use Fix with AI, or roll back to a stable version.

FAQ

Can I use Weaver without writing code?
Yes. Weaver builds workflows entirely from chat. For fine control, refine them in the Studio Canvas.

How is Weaver different from Canvas?
Weaver is built into Studio as a tab. You chat on the left, and the Canvas updates live on the right. Canvas gives you manual control; Weaver speeds up the start.

What if Weaver builds the wrong steps?
Edit or reject them in chat, adjust in Canvas, or follow Debugging Agents.

Can I really build from images or URLs?
Yes. Weaver can understand prompts, API specs, and diagrams—letting you choose whichever fits your workflow best.

Next steps

• Go through the Weaver Quickstart for a guided build.
• Explore prebuilt templates to see ready-to-run examples.
• Learn the available components so you know your building blocks.
• Switch to the Canvas for detailed edits and advanced workflows.
• Deploy through Runtime once your agent is ready.
• Keep Debugging Agents handy while refining.