Deepagents is a simple, open source agent harness that implements a few generally useful tools, including planning (prior to task execution), computer access (giving the able access to a shell and a filesystem), and sub-agent delegation (isolated task execution). This is a UI for interacting with deepagents.
Install dependencies and run the app
$ git clone https://github.com/langchain-ai/deepagents-ui.git
$ cd deepagents-ui
$ yarn install
$ yarn devDeploy a deepagent
As an example, see our deepagents quickstart repo for an example and run the deep_research example.
The langgraph.json file has the assistant ID as the key:
"graphs": {
"research": "./agent.py:agent"
},
Kick off the local LangGraph deployment:
$ cd deepagents-quickstarts/deep_research
$ langgraph devYou will see the local LangGraph deployment log to terminal:
╦ ┌─┐┌┐┌┌─┐╔═╗┬─┐┌─┐┌─┐┬ ┬
║ ├─┤││││ ┬║ ╦├┬┘├─┤├─┘├─┤
╩═╝┴ ┴┘└┘└─┘╚═╝┴└─┴ ┴┴ ┴ ┴
- 🚀 API: http://127.0.0.1:2024
- 🎨 Studio UI: https://smith.langchain.com/studio/?baseUrl=http://127.0.0.1:2024
- 📚 API Docs: http://127.0.0.1:2024/docs
...
You can get the Deployment URL and Assistant ID from the terminal output and langgraph.json file, respectively:
- Deployment URL: http://127.0.1:2024
- Assistant ID:
research
Open Deepagents UI at http://localhost:3000 and input the Deployment URL and Assistant ID:
- Deployment URL: The URL for the LangGraph deployment you are connecting to
- Assistant ID: The ID of the assistant or agent you want to use
- [Optional] LangSmith API Key: Your LangSmith API key (format:
lsv2_pt_...). This may be required for accessing deployed LangGraph applications. You can also provide this via theNEXT_PUBLIC_LANGSMITH_API_KEYenvironment variable.
Usagee
You can interact with the deployment via the chat interface and can edit settings at any time by clicking on the Settings button in the header.
As the deepagent runs, you can see its files in LangGraph state.
You can click on any file to view it.
You can optionally set environment variables instead of using the settings dialog:
NEXT_PUBLIC_LANGSMITH_API_KEY="lsv2_xxxx"Note: Settings configured in the UI take precedence over environment variables.
You can run your Deep Agents in Debug Mode, which will execute the agent step by step. This will allow you to re-run the specific steps of the agent. This is intended to be used alongside the optimizer.
You can also turn off Debug Mode to run the full agent end-to-end.
If the term "Deep Agents" is new to you, check out these videos! What are Deep Agents? Implementing Deep Agents
