A lot of repetitive knowledge work happens around documents; whether it’s analyzing financial statements, processing onboarding forms, downstream tasks and automations over invoices or more. Existing tools around document automation are inaccurate, hard to configure, and can only handle relatively simple tasks. Traditional solutions force you to choose between (a) inflexible low-code builders or (b) time-intensive custom development.
Our new LlamaAgents Builder is both easier to use and more customizable than either alternative. It offers a middle path: describe what you need in natural language, and let us generate the appropriate agent workflow with coding agents.
Need to classify financial statements and extract key metrics? Want to split resumes by section and pull structured data? Building a multi-document summarization pipeline? Just explain your workflow, and our agent builder handles the rest - asking clarifying questions, constructing the appropriate agent Workflow , and connecting each step to the right LlamaCloud document processing tools.
Our launch today is in beta, and free for all LlamaCloud users. You can start vibe-coding your document agents today! We’d love to hear from you when you try it out. Drop us a feedback in the UI and let us know what you think, and what new features you’d like to see.
The resulting document agents will inherently be capable of handling documents of any level of complexity (like PDFs with varying layouts, charts, images and so on) and it will use our document processing tools like LlamaParse, LlamaExtract, LlamaSplit and more to help solve tasks.
Let’s take a look at what the new tool offers.
Why Natural Language for Workflow Building?
Traditional low-code tools force you into predefined boxes. You're stuck with their schema templates, their workflow patterns, their extraction logic. The moment your use-case deviates (and it always does), you're either hacking workarounds or starting over with code. And you're also locked into their deployment environment, what your no/low-code solution hosts is what you get.
On the other hand, writing it all in code gives you flexibility, but at a cost. You're building orchestration logic, handling document routing, managing tool integrations, and debugging complex workflows before you've even validated if your approach works. When it comes to deployment, you’re also on your own.
LlamaAgents Builder cuts through these problems, by changing the way you define the problem space. Describing "classify this document into invoice, receipt, or statement, then extract line items from invoices, totals from receipts, and summary metrics from statements" is fundamentally easier than either:
- Clicking through UI wizards trying to configure three separate extraction schemas and conditional routing logic
- Writing a Workflow class with document classification logic, class-based routing, and tool orchestration from scratch
LlamaAgents Builder generates an actual Workflow : our open-source orchestration framework for building use-case specific agents. You can take what it builds, inspect the code, and customize it further. You get the speed of low-code with the flexibility of full programming. Unlike traditional no-code tools, the Builder compiles your natural language description into actual Python code using our open-source Workflows framework. You can define custom logic without learning pre-built nodes, and extend the workflow however you want through code. It's both easier to start and more powerful to extend than existing approaches.
While the Builder generates your workflow, you'll see a visual representation of your complete agent orchestration. This shows exactly how documents flow through classification, routing, and extraction steps, giving you clarity on the logic before deployment.
The Builder provides the resulting agent workflow in code that you can push to your GitHub. The application can then be deployed and hosted on LlamaCloud, or you can take it and serve within your own infrastructure.
Optimized for Document Extraction Use-Cases
LlamaAgents Builder currently focuses on extraction-based document processing workflows:
- Its aim is to create structured data extraction from varying document formats
- It asks clarifying questions about your extraction task
- It will create the relevant LlamaExtract agents in your LlamaCloud account
- Finally, it will create the appropriate
Workflowwith relevant steps and events.
These use-cases currently serve as the foundation over which the Builder customizes to your specific requirements. The agent can adapt workflows to your particular document types, field structures, and extraction needs. Following today's release, we'll expand the use cases LlamaAgents Builder handles.
Currently, workflows that diverge significantly from extraction patterns may not work as well in this beta. If you're tackling different document workflow types, we want to hear from you—your feedback directly influences what we build next.
Simple UI for Fast Validation
We've kept the beta UI intentionally minimal, focusing on straightforward extraction and review interfaces. This lets you quickly validate your agent's behavior without unnecessary complexity. The focus is on getting from idea to working agent as fast as possible.
Deployment Details
One of the most powerful aspects of LlamaAgents Builder is that you go from conversation to deployed, production-ready agent.
When you click Deploy, the Builder:
- Pushes your generated workflow code to a GitHub repository
- Deploys it to LlamaCloud's managed infrastructure
- Provisions API endpoints for your workflow
- Includes a web UI for uploading documents and reviewing results
The deployed code is yours, under your control. The generated workflow is real Python code using our open-source Workflows framework. You can clone the repository and inspect every line. You can customize the logic beyond what the chat interface provides, push changes and redeploy. You can take the code and deploy on your own infrastructure to meet your organization’s data and privacy requirements.
This is the key difference from no-code tools: you're not locked in. The Builder gives you a fast starting point, but you own the code and can take it anywhere.
Try It and Share Your Feedback
LlamaAgents Builder is live in beta today for all LlamaCloud users. Try it out and drop us feedback through the UI. We're especially interested in hearing about:
- Document workflows beyond extraction patterns you'd like to build
- Use-cases where the current templates don't quite fit
- Features that would make the Builder more useful for your work
Your input shapes what we add next. Signup to LlamaCloud to get started.
