- Overview
- Getting Started with UiPath Agents
- Getting Started with UiPath Agents using LangGraph
- Introduction
- Set up your environment
- Build the agent
- Evaluate the agent
- Connect to Studio Web
- Building a Low-Code Agent in Studio Web
- Adding Tools to Your UiPath Agent
Connect your local LangGraph project to UiPath Studio Web and push your first version.
With local evaluation results confirmed, you are ready to connect the project to Studio Web.
Step 9 - Connect to Studio Web
Your agent is built and evaluated locally. Now connect it to Studio Web so you get version history, evaluation traces, and the ability to test in the cloud UI.
You have three options:
| Option | What happens |
|---|---|
| A: You set it up in Studio Web | Open Studio Web, create a Coded Agent project, copy the project ID. You paste it into .env and push from the CLI. |
| B: CLI packages and uploads | CLI packages the agent and publishes it to your personal workspace feed without Studio Web setup. See the uip solution docs for the command reference. |
| C: Local dev server only | Run uip codedagent dev for a local web UI. Nothing is published to the cloud. |
For this lab, use Option A: it shows the connection model that underlies all three options.
Option A - Connect via Studio Web
-
Log in to UiPath Automation Cloud and open Studio Web from the side navigation.
-
Select Create New and select Agent. Choose Coded as the agent type, then select Start Fresh.
-
Studio Web displays a Setup your coded agent panel. Under Sync from your IDE into Studio Web, your
UIPATH_PROJECT_IDappears with a copy button. Copy this value. -
Open the
.envfile in your project root and add the project ID:UIPATH_PROJECT_ID=your-project-id-hereUIPATH_PROJECT_ID=your-project-id-hereNote:Only the project ID goes in
.env.uip loginstores your auth token globally; you do not needUIPATH_URLorUIPATH_ACCESS_TOKENin.env. If you have used the UiPath Python SDK before and are used to runninguipath authto populate those fields, that step is no longer needed. -
Add
.envto.gitignoreto avoid committing it:echo ".env" >> .gitignoreecho ".env" >> .gitignore -
Push your agent to Studio Web:
uip codedagent pushuip codedagent pushA successful push returns confirmation and increments the version to
0.0.1. Open your project in Studio Web; the agent definition, your evaluation sets, and version0.0.1appear in the version history.
Step 10 - Iterate and improve (optional)
With evaluations in place and Studio Web connected, you can iterate on your agent and observe the effect on scores:
- Review evaluation results in
eval-results.jsonor in Studio Web. - Ask your coding agent to improve the agent based on the evaluation feedback.
- Re-run evaluations to verify improvements.
- Push the updated version to Studio Web.
uip codedagent eval agent evaluations/eval-sets/smoke-test.json --workers 3
uip codedagent eval agent evaluations/eval-sets/smoke-test.json --workers 3
uip codedagent push
uip codedagent push
Each push increments the version in Studio Web, giving you a full history of how the agent evolved.
What you built
You have built a LangGraph agent on UiPath using the CLI and coding agent skills:
- Scaffolded a local Python project with the correct LangGraph structure using
uip codedagent new. - Used your coding agent to build the agent logic, guided by UiPath skills, not detailed prompts.
- Ran the agent locally and verified outputs before touching the cloud.
- Created an evaluation set and scored the agent locally.
- Connected to Studio Web and pushed the first version with full evaluation history.
What's next
- UiPath LangGraph sample agents - working examples including ticket classification with human-in-the-loop, RAG, multi-agent supervisors, and MCP integration.
- UiPath Python SDK docs - full reference for CLI commands, agent patterns, and SDK APIs.
- Evaluation framework guide - how to build, run, and interpret evaluation sets.
- LangGraph docs - the orchestration framework used in this lab.
- UiPath Community - forums, how-tos, and developer discussion.