Deep Research Assistant
Conducts comprehensive research on any topic by searching the web, analyzing sources, and compiling findings into structured reports. Saves research to Notion or Google Docs with citations and summaries.
Toolkits Needed
You'll need access to these toolkits in your Arcade MCP Gateway:
Setup Guide
Sign up for Arcade
Create a free Arcade account to get access to MCP Gateway.
Create Arcade Account →Create a new MCP Gateway
Navigate to MCP Gateway page in Arcade, then click Create MCP Gateway.
Provide a slug that becomes your MCP URL (this is what you'll configure in Agent Builder):
my-agent, your MCP URL will be:https://api.arcade.dev/mcp/my-agent Select the specific tools you need
Arcade MCP Gateway gives you fine-grained control over exactly which tools your agent can access. Select these specific tools from the toolkits:
Why this matters: Instead of granting access to entire toolkits with hundreds of tools, you're selecting only the 11 specific tools this agent actually needs. This precision reduces risk and keeps your agent focused.
Create the gateway and copy your MCP URL
Click Create Gateway. Copy the MCP URL — this is all you need to connect to LangSmith!
https://api.arcade.dev/mcp/my-agent No tokens needed! Arcade uses OAuth for secure authentication — you'll sign in with your Arcade account in LangSmith.
Open LangSmith Agent Builder
This agent is pre-created in the LangSmith Agent Builder templates. Go to LangSmith, find this template, and start using it right away!
Open LangSmith Agent Builder →Can't find this template? Click here for the system prompt
If you don't see this template in your account, you can create the agent manually. Copy this system prompt and paste it into the "System Prompt" field when creating a new agent:
You are a Deep Research Assistant that conducts comprehensive research. ## Your Capabilities - Search Google for information - Scrape websites for content - Read documents in Google Docs - Create research reports in Google Docs - Share findings via Slack ## Workflow 1. **Define** - Understand research question 2. **Search** - Use Google to find sources 3. **Scrape** - Read relevant web pages 4. **Synthesize** - Compile findings 5. **Report** - Create research document ## Example Interactions **User:** "Research the MCP protocol and its adoption" **Agent:** I'll conduct comprehensive research on MCP... **Research Process:** - Searched Google for "MCP protocol AI" - Found 15 relevant sources - Scraped 8 key articles **Key Findings:** 1. MCP (Model Context Protocol) is... 2. Major adopters include... 3. Use cases range from... Full report created in Google Docs. Want me to share on Slack? ## Guidelines - Use multiple sources - Verify information - Cite sources properly - Synthesize, don't just summarize
Connect your Arcade MCP Gateway
In the Agent Builder, click the MCP button and paste your MCP URL from Step 4:
Click Connect to initiate the OAuth flow:
Click Allow to grant LangSmith access to your Arcade tools:
Select the tools for your agent
After connecting, click the Add tools button. Your gateway exposes only the specific tools you configured:
Start using your agent!
That's it! Your agent is now connected to Arcade's tools. Start chatting with your agent to test it out.
Try It: Example Prompt
Once your agent is set up, try this example prompt to test it:
Research "AI agent frameworks" - search Google, read the top articles, and create a comprehensive report in Google Docs covering the main frameworks, their pros/cons, and adoption trends. Share the report on #research when done.
Best For
