Skip to main content
Glama

Browser-Use MCP Server

.env.example6 kB
# This is an example .env file. Copy it to .env and fill in your values. # Lines starting with # are comments. # === Main LLM Configuration (MCP_LLM_*) === # Select the primary LLM provider # Options: openai, azure_openai, anthropic, google, mistral, ollama, deepseek, openrouter, alibaba, moonshot, unbound MCP_LLM_PROVIDER=google # Specify the model name for the selected provider MCP_LLM_MODEL_NAME=gemini-2.5-flash-preview-04-17 # LLM temperature (0.0-2.0). Controls randomness. MCP_LLM_TEMPERATURE=0.0 # Optional: Generic override for the LLM provider's base URL # MCP_LLM_BASE_URL= # Optional: Generic override for the LLM provider's API key. Takes precedence over provider-specific keys. # MCP_LLM_API_KEY= # --- Provider Specific API Keys (MCP_LLM_*) --- # Required unless using Ollama locally without auth or generic MCP_LLM_API_KEY is set # MCP_LLM_OPENAI_API_KEY=YOUR_OPENAI_API_KEY # MCP_LLM_ANTHROPIC_API_KEY=YOUR_ANTHROPIC_API_KEY # MCP_LLM_GOOGLE_API_KEY=YOUR_GOOGLE_API_KEY # MCP_LLM_AZURE_OPENAI_API_KEY=YOUR_AZURE_OPENAI_API_KEY # MCP_LLM_DEEPSEEK_API_KEY=YOUR_DEEPSEEK_API_KEY # MCP_LLM_MISTRAL_API_KEY=YOUR_MISTRAL_API_KEY # MCP_LLM_OPENROUTER_API_KEY=YOUR_OPENROUTER_API_KEY # MCP_LLM_ALIBABA_API_KEY=YOUR_ALIBABA_API_KEY # MCP_LLM_MOONSHOT_API_KEY=YOUR_MOONSHOT_API_KEY # MCP_LLM_UNBOUND_API_KEY=YOUR_UNBOUND_API_KEY # --- Provider Specific Endpoints (MCP_LLM_*) --- # Optional: Override default API endpoints. # MCP_LLM_OPENAI_ENDPOINT=https://api.openai.com/v1 # MCP_LLM_ANTHROPIC_ENDPOINT=https://api.anthropic.com # MCP_LLM_AZURE_OPENAI_ENDPOINT=YOUR_AZURE_ENDPOINT # Required if using Azure, e.g., https://your-resource.openai.azure.com/ # MCP_LLM_AZURE_OPENAI_API_VERSION=2025-01-01-preview # MCP_LLM_DEEPSEEK_ENDPOINT=https://api.deepseek.com # MCP_LLM_MISTRAL_ENDPOINT=https://api.mistral.ai/v1 # MCP_LLM_OLLAMA_ENDPOINT=http://localhost:11434 # MCP_LLM_OPENROUTER_ENDPOINT=https://openrouter.ai/api/v1 # MCP_LLM_ALIBABA_ENDPOINT=https://dashscope.aliyuncs.com/compatible-mode/v1 # MCP_LLM_MOONSHOT_ENDPOINT=https://api.moonshot.cn/v1 # MCP_LLM_UNBOUND_ENDPOINT=https://api.getunbound.ai # --- Ollama Specific (MCP_LLM_*) --- # MCP_LLM_OLLAMA_NUM_CTX=32000 # MCP_LLM_OLLAMA_NUM_PREDICT=1024 # === Planner LLM Configuration (Optional, MCP_LLM_PLANNER_*) === # If you want to use a different LLM for planning tasks within agents. # Defaults to main LLM settings if not specified. # MCP_LLM_PLANNER_PROVIDER= # MCP_LLM_PLANNER_MODEL_NAME= # MCP_LLM_PLANNER_TEMPERATURE= # MCP_LLM_PLANNER_BASE_URL= # MCP_LLM_PLANNER_API_KEY= # Generic planner API key, or use provider-specific below # MCP_LLM_PLANNER_OPENAI_API_KEY= # ... (similar provider-specific keys and endpoints for planner if needed) # === Browser Configuration (MCP_BROWSER_*) === # General browser headless mode (true/false) MCP_BROWSER_HEADLESS=false # General browser disable security features (use cautiously) (true/false) MCP_BROWSER_DISABLE_SECURITY=false # Optional: Path to Chrome/Chromium executable # MCP_BROWSER_BINARY_PATH=/usr/bin/chromium-browser # Optional: Path to Chrome user data directory (for persistent sessions) # MCP_BROWSER_USER_DATA_DIR=~/.config/google-chrome/Profile 1 MCP_BROWSER_WINDOW_WIDTH=1280 MCP_BROWSER_WINDOW_HEIGHT=1080 # Set to true to connect to user's browser via MCP_BROWSER_CDP_URL MCP_BROWSER_USE_OWN_BROWSER=false # Optional: Connect to existing Chrome via DevTools Protocol URL. Required if MCP_BROWSER_USE_OWN_BROWSER=true. # MCP_BROWSER_CDP_URL=http://localhost:9222 # MCP_BROWSER_WSS_URL= # Optional: WSS URL if CDP URL is not sufficient # Keep browser managed by server open between MCP tool calls (if MCP_BROWSER_USE_OWN_BROWSER=false) MCP_BROWSER_KEEP_OPEN=false # Optional: Directory to save Playwright trace files (useful for debugging). If not set, tracing to file is disabled. # MCP_BROWSER_TRACE_PATH=./tmp/trace # === Agent Tool Configuration (`run_browser_agent` tool, MCP_AGENT_TOOL_*) === MCP_AGENT_TOOL_MAX_STEPS=100 MCP_AGENT_TOOL_MAX_ACTIONS_PER_STEP=5 # Method for tool invocation ('auto', 'json_schema', 'function_calling') MCP_AGENT_TOOL_TOOL_CALLING_METHOD=auto MCP_AGENT_TOOL_MAX_INPUT_TOKENS=128000 # Enable vision capabilities (screenshot analysis) MCP_AGENT_TOOL_USE_VISION=true # Override general browser headless mode for this tool (true/false/empty for general setting) # MCP_AGENT_TOOL_HEADLESS= # Override general browser disable security for this tool (true/false/empty for general setting) # MCP_AGENT_TOOL_DISABLE_SECURITY= # Enable Playwright video recording (true/false) MCP_AGENT_TOOL_ENABLE_RECORDING=false # Optional: Path to save agent run video recordings. If not set, recording to file is disabled even if ENABLE_RECORDING=true. # MCP_AGENT_TOOL_SAVE_RECORDING_PATH=./tmp/recordings # Optional: Directory to save agent history JSON files. If not set, history saving is disabled. # MCP_AGENT_TOOL_HISTORY_PATH=./tmp/agent_history # === Deep Research Tool Configuration (`run_deep_research` tool, MCP_RESEARCH_TOOL_*) === MCP_RESEARCH_TOOL_MAX_PARALLEL_BROWSERS=3 # MANDATORY: Base directory to save research artifacts (report, results). Task ID will be appended. # Example: MCP_RESEARCH_TOOL_SAVE_DIR=/mnt/data/research_outputs # Example: MCP_RESEARCH_TOOL_SAVE_DIR=C:\\Users\\YourUser\\Documents\\ResearchData MCP_RESEARCH_TOOL_SAVE_DIR=./tmp/deep_research # === Path Configuration (MCP_PATHS_*) === # Optional: Directory for downloaded files. If not set, persistent downloads to a specific path are disabled. # MCP_PATHS_DOWNLOADS=./tmp/downloads # === Server Configuration (MCP_SERVER_*) === # Path for the server log file. Leave empty for stdout. # MCP_SERVER_LOG_FILE=mcp_server_browser_use.log # Logging level: DEBUG, INFO, WARNING, ERROR, CRITICAL MCP_SERVER_LOGGING_LEVEL=INFO # Enable/disable anonymized telemetry (true/false) MCP_SERVER_ANONYMIZED_TELEMETRY=true # Optional: JSON string for MCP client configuration for the controller # MCP_SERVER_MCP_CONFIG='{"client_name": "mcp-browser-use-controller"}'

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/Saik0s/mcp-browser-use'

If you have feedback or need assistance with the MCP directory API, please join our Discord server