Run an agent and toolkit at the Room level for testing.
  1. Set Environment Variables
    The following environment variables must be set to configure Firecrawl and handle API access, retries, and rate limits:
    • FIRECRAWL_API_URL – API base URL (default: https://api.firecrawl.dev/v1)
    • FIRECRAWL_RETRY_MAX_ATTEMPTS – Max retry attempts (default: 5)
    • FIRECRAWL_RETRY_INITIAL_DELAY – Initial retry delay in ms (default: 2000)
    • FIRECRAWL_RETRY_MAX_DELAY – Max retry delay in ms (default: 30000)
    • FIRECRAWL_RETRY_BACKOFF_FACTOR – Exponential backoff multiplier (default: 3)
    • FIRECRAWL_CREDIT_WARNING_THRESHOLD – Credit warning threshold (default: 2000)
    • FIRECRAWL_CREDIT_CRITICAL_THRESHOLD – Credit critical threshold (default: 500)
    • FIRECRAWL_API_KEY – Your Firecrawl API key (required)
  2. Install Meshagent CLI 
    pip install "meshagent[all]"
  3. Sign Up & Authenticate with Meshagent
    See: https://docs.meshagent.com/cli/getting_started
  4. Start MCP Agent Service in a Test Room 
    meshagent service test \
  --room=test \
  --role=agent \
  --image=meshagent/mcp-firecrawl:latest \
  --env=MESHAGENT_PORT=8001 \
  --env=FIRECRAWL_API_URL=https://api.firecrawl.dev/v1 \
  --env=FIRECRAWL_RETRY_MAX_ATTEMPTS=5 \
  --env=FIRECRAWL_RETRY_INITIAL_DELAY=2000 \
  --env=FIRECRAWL_RETRY_MAX_DELAY=30000 \
  --env=FIRECRAWL_RETRY_BACKOFF_FACTOR=3 \
  --env=FIRECRAWL_CREDIT_WARNING_THRESHOLD=2000 \
  --env=FIRECRAWL_CREDIT_CRITICAL_THRESHOLD=500 \
  --env=FIRECRAWL_API_KEY=YOUR-API-KEY \
  --port="num=8001 path=/webhook liveness=/ type=meshagent.callable" \
  --name=mcp-firecrawl-service-test
    • Starts a Meshagent Room with a Firecrawl MCP server agent available. Room and all its agents are automatically removed if idle.
  5. Join a Chatbot with Firecrawl Toolkit 
    meshagent chatbot join \
  --room=test \
  --agent-name=mcp-firecrawl \
  --name=mcp-firecrawl \
  --toolkit=mcp-firecrawl
    • Starts a chatbot in the Room using the Firecrawl toolkit. Multiple toolkits and agents can participate in the same Room.
    • Output will show a link to the new Room.
  6. Test in the Browser
    • Open the Room link in your browser and send a message to the agent to interact with the Firecrawl MCP Server tools.

​
Project Level Deployment

For production, deploy MCP Server tools and chatbot persistently at the project level so they auto-join every Room created in the project.
  1. Create Persistent MCP Agent Service 
    meshagent service create \
  --role=agent \
  --image=meshagent/mcp-firecrawl:latest \
  --env=MESHAGENT_PORT=8001 \
  --env=FIRECRAWL_API_URL=https://api.firecrawl.dev/v1 \
  --env=FIRECRAWL_RETRY_MAX_ATTEMPTS=5 \
  --env=FIRECRAWL_RETRY_INITIAL_DELAY=2000 \
  --env=FIRECRAWL_RETRY_MAX_DELAY=30000 \
  --env=FIRECRAWL_RETRY_BACKOFF_FACTOR=3 \
  --env=FIRECRAWL_CREDIT_WARNING_THRESHOLD=2000 \
  --env=FIRECRAWL_CREDIT_CRITICAL_THRESHOLD=500 \
  --env=FIRECRAWL_API_KEY=YOUR-API-KEY \
  --port="num=8001 path=/webhook liveness=/ type=meshagent.callable" \
  --name=mcp-firecrawl-service
  2. Create Persistent Chatbot Service 
    meshagent service create \
  --image="meshagent/cli:latest" \
  --port="num=9001 path=/agent liveness=/ type=meshagent.callable participant_name=mcp-firecrawl-chatbot" \
  --env="MESHAGENT_PORT=9001" \
  --name="mcp-firecrawl-chatbot-service" \
  --command="meshagent chatbot service --agent-name=mcp-firecrawl-chatbot --toolkit=mcp-firecrawl"
This ensures the Firecrawl MCP server and chatbot are present in every Room in your Meshagent project, with no need to manage local or separate instances.

​
Tools Available

The following tools are available via the Firecrawl MCP Server in your Meshagent Room:

​
Tool: firecrawl_check_crawl_status

Check the status of a crawl job.
  • Returns: Status and progress of the crawl job, including results if available.

​
Tool: firecrawl_crawl

Starts an asynchronous crawl job on a website and extracts content from all pages.
  • Best for: Extracting content from multiple related pages for comprehensive coverage.
  • Returns: Operation ID for status checking.

​
Tool: firecrawl_deep_research

Conduct deep web research on a query using intelligent crawling, search, and LLM analysis.
  • Best for: Complex research questions requiring in-depth analysis from multiple sources.
  • Returns: LLM-generated analysis and sources used.

​
Tool: firecrawl_extract

Extract structured information from web pages using LLM capabilities.
  • Best for: Extracting specific structured data like prices, product details.
  • Returns: Extracted structured data as defined by a JSON schema.

​
Tool: firecrawl_generate_llmstxt

Generate a standardized llms.txt (and optionally llms-full.txt) file for a domain.
  • Best for: Creating machine-readable permission guidelines for AI models.

​
Tool: firecrawl_map

Map a website to discover all indexed URLs on the site.
  • Best for: Discovering URLs before deciding what to scrape.

​
Tool: firecrawl_scrape

Scrape content from a single URL with advanced options.
  • Best for: Single page content extraction, when you know the exact page.
Search the web and optionally extract content from search results.
  • Best for: Finding specific information across multiple sites.
See the tool details above for complete parameters and usage.