OpenManus Web Version is here

Written by
Iris Vance
Updated on:July-03rd-2025
Recommendation

No command line required, OpenManus Web version brings a new experience.

Core content:
1. OpenManus Web version usage guide
2. Detailed steps for configuring large models
3. Run the Web version and query problems

Yang Fangxian
Founder of 53AI/Most Valuable Expert of Tencent Cloud (TVP)

If you don't like command line operationOpenManus, just try itWebVersionOpenManusBar!

Download the front-end branch

First, downloadOpenManusoffront-endBranch code, download address: https://github.com/mannaandpoem/OpenManus/tree/front-end

Configuring large models

andmainLike branches, you need toconfig.tomlMedium Configurationkey, decompressOpenManus-front-endPackage, enter the directory:

cd OpenManus-front-end/config
cp config.example.toml config.toml

For example, I configuredDeepSeekLarge Model:

#  Global LLM configuration
[llm]
model = "deepseek-chat"
base_url = "https://api.deepseek.com"
api_key = "sk-xxxxxxxxxxxx"
max_tokens = 4096
temperature = 0.0

#
 Optional configuration  for  specific LLM models
[llm.vision]
model = "deepseek-chat"
base_url = "https://api.deepseek.com"
api_key = "sk-xxxxxxxxxxxx"

#
 Server configuration
[server]
host = "localhost"
port = 5172

After configuration, run:

python app.py

This interface runs on the localhttp://localhost:5172You don't need to manually access this address, the front-end interface will pop up automatically: At this time, you can enter questions in the dialog box, such as: Check the weather in Guiyang today.

Switch search engine

Deadline2025Year3moon15day10When this branch is not yet likemainWhen merging other search engines like branches, the following error will occur when using the browser to perform tasks:

2025-03-15 10:14:34.701 | ERROR | app.agent.toolcall:execute_tool:166 - ⚠️ Tool 'google_search' encountered a problem: HTTPSConnectionPool(host='www.google.com', port=443): Max retries exceeded with url: /search?q=%E8%B4%B5%E9%98%B3+%E4%BB%8A%E5%A4%A9+%E5%A4%A9%E6%B0%94&num=3&hl=en&start=0&safe=active (Caused by ConnectTimeoutError(<urllib3.connection.HTTPSConnection object at 0x10af4c050>, 'Connection to www.google.com timed out. (connect timeout=5)'))
2025-03-15 10:14:34.702 | INFO | app.agent.toolcall:act:114 - ? Tool 'google_search' completed its mission! Result: Error: ⚠️ Tool 'google_search' encountered a problem: HTTPSConnectionPool(host='www.google.com', port=443): Max retries exceeded with url: /search?q=%E8%B4%B5%E9%98%B3+%E4%BB%8A%E5%A4%A9+%E5%A4%A9%E6%B0%94&num=3&hl=en&start=0&safe=active (Caused by ConnectTimeoutError(<urllib3.connection.HTTPSConnection object at 0x10af4c050>, 'Connection to www.google.com timed out. (connect timeout=5)'))

However, according toOpenManusConfigure Baidu search engine according to the original solution in Switch Search Engine. After configuration, re-run:

python app.py

There will be a log in the background prompting the use of Baidu search engine:

2025-03-15 10:18:38.397 | INFO | app.agent.toolcall:think:54 - ✨ Manus's thoughts: 
2025-03-15 10:18:38.398 | INFO | app.agent.toolcall:think:55 - ?️ Manus selected 1 tools to use
2025-03-15 10:18:38.398 | INFO | app.agent.toolcall:think:59 - ? Tools being prepared: ['baidu_search']
2025-03-15 10:18:38.398 | INFO | app.agent.toolcall:execute_tool:144 - ? Activating tool: 'baidu_search'...
INFO: ::1:51037 - "GET /tasks/b986ce2a-6f3b-4e7b-82dc-855e3c4baa62 HTTP/1.1" 200 OK
INFO: ::1:51038 - "GET /tasks/b986ce2a-6f3b-4e7b-82dc-855e3c4baa62 HTTP/1.1" 200 OK
INFO: ::1:51037 - "GET /tasks/b986ce2a-6f3b-4e7b-82dc-855e3c4baa62 HTTP/1.1" 200 OK
INFO: ::1:51038 - "GET /tasks/b986ce2a-6f3b-4e7b-82dc-855e3c4baa62 HTTP/1.1" 200 OK
2025-03-15 10:18:39.845 | INFO | app.agent.toolcall:act:114 - ? Tool 'baidu_search' completed its mission! Result: Observed output of cmd `baidu_search` executed:

Of course, the front end will also display: Configure the Baidu search engine to "check Guiyang's weather today":

User experience

Personal feelingOpenManusSince there iswebVersion, some functions can be added appropriately, such as configuring large models to be able towebInterface execution, the task can be terminated during execution, etc. It has just grown and will definitely be more perfect in the future.