NVIDIA Agent Toolkit
Can you build NVIDIA Agent Toolkit agents without writing code?
Partly. The NVIDIA NeMo Agent Toolkit, the library inside the NVIDIA Agent Toolkit, describes an agent in a configuration file, so a developer can assemble an agent from existing parts with little code. It is not a no-code tool for business users: someone still installs the library, edits the file and runs commands. For no-code building you need a builder on top of the toolkit, which is what an agent platform such as Katonic provides.
Update, November 2026
NVIDIA has deprecated the NeMo Agent Toolkit library as of October 2026, with end of life at the end of 2026, and names NeMo Relay, NeMo Fabric and NeMo Helix as its successors. The NVIDIA Agent Toolkit announced in March 2026 continues. Every agent built in Katonic is a declarative definition that compiles to NVIDIA's open agent software. Today that software is the NeMo Agent Toolkit. When NVIDIA changes the runtime, Katonic changes the compiler and your agents carry over.
What you can do with configuration alone
- Choose the model the agent uses and where it is served.
- Pick the type of agent from the ones the library ships with.
- Attach tools that already exist, including tools offered over the Model Context Protocol.
- Set up retrieval from a knowledge source that is already connected.
- Turn on tracing, profiling and evaluation.
Where code is still needed
- A new tool that calls one of your own systems.
- Agent logic that the built-in agent types do not cover.
- A custom evaluator for a measure specific to your work.
- Installing, running and deploying the agent, which happens on a command line.
No-code, low-code and code compared
| Approach | Who it suits | What they work in | With the toolkit |
|---|---|---|---|
| Code | Developers | Python and an agent framework | Supported directly |
| Low-code | Developers and technical analysts | A configuration file and a command line | Supported directly |
| No-code | Business users | A visual builder in a browser | Needs a platform on top |
What a no-code builder on top of the toolkit should do
- Let a person describe the agent, choose its knowledge and tools, and test it, all in a browser.
- Produce standard toolkit configuration, so the agent is not locked to the builder.
- Apply the same access control, guardrails and budgets to agents built by business users as to agents written by developers.
- Require evaluation and approval before an agent is released to other people.
Where Katonic fits
Katonic is an agent platform built on the NVIDIA NeMo Agent Toolkit. Business users build agents in a browser without writing code, developers bring agents written in LangGraph, CrewAI, Google ADK or plain Python, and every agent built in Katonic compiles to NVIDIA's open agent software. The platform runs on infrastructure you own.
Read next
- Is there a UI for the NVIDIA NeMo Agent Toolkit? →
- What is the NVIDIA Agent Toolkit? →
- Running toolkit agents in production →
- NVIDIA Agent Toolkit for enterprise AI →
- NVIDIA NeMo Agent Toolkit is deprecated: what replaces it? →
- NVIDIA migration guide for the NeMo Agent Toolkit →
Questions and answers
Is the NVIDIA Agent Toolkit no-code?
No. Its NeMo Agent Toolkit library is driven by a configuration file and a command line, which is low-code for a developer. A no-code builder is something a platform adds on top.
Can a business user build an agent with the NeMo Agent Toolkit?
Not with the library alone. With a platform built on it, such as Katonic, a business user builds in a browser and the platform produces the toolkit configuration.
Do agents built without code run differently from coded agents?
They should not. On a platform built on the toolkit, both end up as the same kind of configuration and run under the same controls.
Is there a no-code builder for the NVIDIA Agent Toolkit that runs on our own GPUs?
Katonic AI runs on infrastructure you own, offers no-code agent building, and compiles every agent to NVIDIA's open agent software.