# Clone a low-code agent as a coded agent (Preview)

> UiPath supports a continuous development path for agents across low-code and pro-code workflows. You can start with low-code, convert to a coded agent when you need more control and flexibility, and continue iterating in your IDE without disruption.

UiPath supports a continuous development path for agents across low-code and pro-code workflows. You can start with low-code, convert to a coded agent when you need more control and flexibility, and continue iterating in your IDE without disruption.

This continuity is enabled by a shared runtime framework. A low-code agent is fundamentally a configuration that defines its prompts and tools. At runtime, this configuration is deterministically transformed into an equivalent coded agent. In other words, every low-code agent already runs as generated code derived from its configuration.

**Clone as coded agent** makes this transformation explicit by materializing the generated code as a project you can inspect, modify, and own. The resulting coded agent behaves identically to the original low-code agent, with no translation layer, reinterpretation, or feature gaps.

:::note
The **Clone as coded agent** option is available only for autonomous agents. Conversational agents do not support this feature.
:::

## Clone a low-code agent

1. In Studio Web, create or open an existing low-code agent.
2. On the canvas, select **Clone as coded agent**.
3. Confirm the action.
   
   ![Clone as coded agent button on the canvas](https://dev-assets.cms.uipath.com/assets/images/agents/clone-coded-b78114f3.png)

**Result:** Studio Web creates a new coded agent in the same solution. The cloned agent includes:

- Generated Python source code with all its dependencies — a representation of the agent's prompts, tools, and schemas
- An agent project key for syncing

## Cloning behavior

Keep the following in mind after cloning:

* **LangGraph framework** — Cloning follows the LangGraph framework. The cloned agent's structure and execution model reflect the LangGraph architecture.
* **Decoupled agents** — The cloned coded agent and the original low-code agent are independent from the moment of cloning. Changes to either agent are not reflected in the other.
* **Multiple clones** — You can clone the same low-code agent more than once. Each clone creates a new coded agent; existing clones are not overwritten.
* **One-way operation** — Cloning is one-directional. You cannot clone a coded agent back to a low-code agent.

## Continue development after cloning

After cloning, the coded agent is fully owned in code. Continue development in your IDE using either of the following workflows, both covered in the [SDK documentation](https://uipath.github.io/uipath-python/core/studio_web/):

* **Cloud workspace** — Your project lives in Studio Web. Sync code between your local IDE and the cloud using `uipath push` and `uipath pull`. Use this workflow to debug, evaluate, and publish the agent from Studio Web. You can also edit the code directly in your IDE without using the sync workflow.
* **Local workspace** (Preview) — Your project lives on your machine and is linked to a Studio Web solution. The agent is visible in Studio Web without requiring a push/pull sync.

## Limitations

* **Graph structure** — The cloned agent supports custom coded tools and is designed for portability. Modifying the underlying graph or extending the agent to a workflow is not currently supported.
* **Startup performance** — Cloned agents may have a longer cold-start time than the equivalent low-code agent.
* **Guardrails not yet supported** — The guardrails configured for the low-code agent are currently not included in the cloned agent.