Build Agents with Weaver

Weaver is the prompt-first interface in SmythOS that lets you build AI agents by simply describing what you want. Unlike traditional agent builder tools that ask you to think like a developer, Weaver is task-oriented: you describe goals; Weaver builds.

What is Weaver?

Weaver lets you describe your automation goal in natural language. It interprets your prompt, builds the logic, and delivers a working agent ready to run.

Weaver transforms your text instructions into complete, testable workflows without needing to write a single line of code.
It’s where your prompt becomes agent... and your typing becomes Smyth-ing.

Understand Agent Weaver in 3 Minutes

While Studio gives you a classic drag-and-drop canvas for building agents manually, Weaver starts with natural conversation in a single prompt box.

Just describe what you need. Weaver thinks, plans, and builds it for you; piece by piece.

Here’s how it works:

1. You describe your goal
   e.g. "Help me write SEO-optimized articles for my blog about calculators"

2. Weaver interprets your intent

3. It asks clarifying questions
   To tailor the agent to your use case

4. It proposes a workflow plan

   • Generate blog structure
   • Write content
   • Research topics
   • Publish (optional)

5. It auto-creates the workflow
   Using components from the SmythOS library

6. You test, tweak, and deploy
   All within the chat

Create a research agent that finds product reviews and summarizes them in markdown.

Want a head start?

Check the available prebuilt templates to see what’s possible, then remix to fit your use case.

Key Features of Weaver

Capability	Description
Prompt-Based Build	Start with plain text. Weaver handles planning and assembly.
Auto-Linked Components	It picks components and wires them based on data needs.
Live Testing	Try agents directly in the chat window.
AI-Powered Fixes	Use “Fix with AI” to repair misconfigured blocks.
Zero-Code Deployment	Deploy to API or Web UI instantly.

How to Get Started with Weaver

We call this flow Chat-to-Agent:

1. Describe the agent
   Create an agent that writes SEO articles and posts to WordPress.
2. Optionally list required skills
   The agent should have the following skills... It should write SEO-optimised articles, extract primary and secondary keywords, and post content to WordPress.
3. Answer smart follow-ups
   Weaver may ask clarifying questions like:
   • What kind of blogs are you targeting?
   • Do you want keyword extraction to be optional or required?
   • Which CMS platform are you using — WordPress.com or self-hosted?
4. Watch Weaver compose the logic
   It picks the right components, configures them, and connects them automatically.
5. Test, fix, and go live

Weaver also supports image uploads or diagrams if you prefer visual instruction.

Fixing with AI

When something breaks, Weaver helps you debug efficiently:

• See input/output previews
• Use “Fix with AI” to auto-correct components
• Select and fix multiple components at once (Cmd/Ctrl + click)

You're in control with Weaver

Weaver never rebuilds the entire agent unless you explicitly request it.

Fix with AI

Component broken? Use the contextual repair tool to edit only what’s broken, not the entire flow.

Working with AI-Suggested Modifications

When Weaver suggests changes or adds new components, keep these best practices in mind to avoid unexpected errors:

1. Verify Component Capabilities

• Weaver may recommend parameters or features that don’t actually exist in the SmythOS component
• Always cross-reference the configuration against the Component Library

2. Understand Component Limitations

• Components have defined scopes and cannot perform tasks outside their built-in functionality
• If a suggestion sounds logical but fails, check if the capability is actually supported

3. Troubleshooting Broken Configurations

• Compare the broken component with the official schema
• Remove any undocumented inputs or outputs
• Rebuild from default if needed, testing one change at a time

4. Use Alternative Paths

• If a feature doesn’t exist, ask Weaver for another way to achieve your goal
• Often, the same outcome can be reached with a different approach using supported tools

5. Test Step-by-Step

• Run each modified block individually before testing the whole agent
• This helps isolate the point of failure

Weaver is smart, but not always accurate

Weaver sometimes suggests features based on logical inference rather than actual implementation. Always confirm with documentation when in doubt.

Inside Weaver

Weaver combines:

• Prompt parsing
• Skill suggestion
• Component wiring
• Live documentation search

It's built to guide, not overwhelm.

Understand Key Concepts of Weaver

Term	Description
Agent	A task-running bot using LLMs and modular skills
Workflow	A sequence of connected components
Component	A unit that performs a function (e.g. API call, content generation)

More on error handling: Debugging Agents

Who Should Use Weaver?

• Ops leads looking to automate processes
• PMs and analysts prototyping AI workflows
• Developers who want a faster starting point
• Anyone who prefers task-based logic over manual configuration

Weaver is for everyone

Whether you're technical or not, Weaver reduces the friction of building intelligent agents.