Content
# Docker + Ollama + Qwen3:8B + LibreChat + Playwright MCP
This project demonstrates how to run a fully local AI-powered browser automation assistant using **LibreChat**, **Ollama**, **Qwen3:8B**, and **Playwright MCP**.
The AI agent can understand natural language instructions, launch a browser, perform actions using Playwright, and summarize the results.
---
# Architecture
```
User
│
▼
LibreChat
│
▼
Qwen3:8B (Ollama)
│
▼
Playwright MCP Server
│
▼
Chrome Browser
```
---
# Prerequisites
## Hardware
* 16 GB RAM or higher
* SSD with at least 20 GB free space
* GPU recommended (optional)
## Software
* Docker Desktop
* Git
* Node.js 20+
* Ollama
---
# Technology Stack
* Docker
* LibreChat
* Ollama
* Qwen3:8B
* Playwright MCP
* Chrome Browser
---
# Step 1 – Install Ollama
Install Ollama from:
https://ollama.com
or using brew
`brew install ollama`
Verify installation:
```bash
ollama --version
```
---
# Step 2 – Download the Model
Pull Qwen3:8B
```bash
ollama pull qwen3:8b
```
Verify:
```bash
ollama list
```
---
# Step 3 – Start Ollama
```bash
ollama serve
```
Default endpoint:
```
http://localhost:11434
```
---
# Step 4 – Clone LibreChat
```bash
git clone https://github.com/danny-avila/LibreChat.git
cd LibreChat
```
---
# Step 5 – Configure LibreChat
Create or edit:
```
librechat.yaml
```
Example MCP configuration:
```yaml
mcpServers:
playwright:
type: sse
url: http://host.docker.internal:8931/sse
timeout: 120000
mcpSettings:
allowedDomains:
- 'host.docker.internal:8931'
- 'localhost:8931'
```
Example Ollama configuration:
```yaml
- name: "Ollama"
apiKey: "ollama"
baseURL: "http://host.docker.internal:11434/v1"
models:
default:
- "qwen3:8b"
fetch: true
titleConvo: true
titleModel: "current_model"
summarize: false
summaryModel: "current_model"
modelDisplayLabel: "Ollama"
```
To verify the connection for Ollama configuration, run the following in the terminal:
`docker exec -it LibreChat curl http://host.docker.internal:11434/api/tags`
Check out the updated project forked in my account [mfaisalkhatri/LibreChat](https://github.com/mfaisalkhatri/LibreChat)
---
# Step 6 – Configure Ollama in LibreChat
Example endpoint
```
http://host.docker.internal:11434
```
Select
* Provider: Ollama
* Model: qwen3:8b
---
# Step 7 – Start Playwright MCP
Open another terminal.
```bash
npx @playwright/mcp@latest \
--host 0.0.0.0 \
--allowed-hosts "*" \
--port 8931 \
```
Server starts at
```
http://localhost:8931
```
---
# Step 8 – Install Chrome
```bash
npx playwright install chrome
```
---
# Step 9 – Start LibreChat
```bash
docker compose up -d
```
Open
```
http://localhost:3080
```
Create your account.
---
# Create an AI Agent using LibreChat
- Add Qwen3:8b model to the Agent
# Recommended Agent Settings
| Setting | Recommended Value |
| ----------------- | ----------------- |
| Temperature | 0.1 |
| Top P | 0.9 |
| Frequency Penalty | 0 |
| Presence Penalty | 0 |
| Reasoning Effort | Medium |
| Reasoning Summary | Auto |
- Add Playwright MCP server to the Agent
- Add the instructions from the `AI-Instructions.txt`
---
# Example Prompts
Open https://playwright.dev and click **Get Started**.
---
Navigate to GitHub and search for Playwright.
---
Open Google and search for "Playwright MCP".
---
Navigate to https://parabank.parasoft.com/parabank/index.htm
Locate "Username" field using "name=username"
Enter "john" into the "Username" field.
Locate "Password" field using "name=password"
Enter "demo" into the "Password" field.
Locator "Log In" button using "input[type="submit"]
Click on the "Log In" button
Verify that the "Accounts Overview" page is displayed
---
# Execution Flow
```
User Prompt
│
▼
LibreChat
│
▼
Qwen3:8B
│
▼
Playwright MCP
│
▼
Chrome
│
▼
Browser Actions
│
▼
Execution Results
│
▼
AI Summary
```
---
# Troubleshooting
## MCP connection closed
Verify the MCP server is running.
```bash
npx @playwright/mcp@latest \
--host 0.0.0.0 \
--allowed-hosts "*" \
--port 8931 \
```
---
## Chrome not found
Install Chrome
```bash
npx playwright install chrome
```
---
## Ollama not detected
Ensure Ollama is running.
```bash
ollama serve
```
Verify:
```
http://localhost:11434
```
---
## LibreChat cannot reach MCP
If LibreChat runs inside Docker, use:
```
host.docker.internal
```
instead of
```
localhost
```
---
## Model is hallucinating
Reduce temperature to:
```
0.2
```
Use stricter instructions for the AI Agent.
---
# Useful Commands
Start Ollama
```bash
ollama serve
```
List models
```bash
ollama list
```
Download model
```bash
ollama pull qwen3:8b
```
Run model
```bash
ollama run qwen3:8b
```
Start Playwright MCP
```bash
npx @playwright/mcp@latest \
--host 0.0.0.0 \
--allowed-hosts "*" \
--port 8931 \
```
Install Chrome
```bash
npx playwright install chrome
```
Start LibreChat
```bash
docker compose up -d
```
Check LibreChat logs in docker
```bash
docker logs Librechat -f
```
Stop LibreChat
```bash
docker compose down
```
