Brave Search + Google Gemini, instant research summaries

Q: How do I adapt this Brave Gemini summaries solution to my specific challenges?

You’ll mainly adjust the system prompt inside the Brave Search Orchestrator agent to control tone, structure, and what “good” looks like. If your MCP provider uses different tool names or endpoints, swap the credentials on Brave Web Lookup and Brave Local Lookup. Common customizations include adding “include sources,” forcing bullet summaries for briefs, and requiring a short “so what” section for decision-makers.

You open 12 tabs to answer one “quick question,” then lose the best source, paste the wrong quote, and still end up with a summary you don’t trust.

This is where Brave Gemini summaries help most. A marketing manager trying to write credible content feels it. A founder doing quick market checks feels it too. Even an agency lead building client briefs gets dragged into the same tab-hopping routine.

This workflow turns one natural-language question into a clean, reusable research summary, using Brave Search for retrieval and Google Gemini for the write-up. You’ll see what it does, what you need, and how to adapt it for your team.

How This Automation Works

See how this solves the problem:

n8n Workflow Template: Brave Search + Google Gemini, instant research summaries

Click to explore

flowchart LR

    subgraph sg0["Brave Search Workflow Start Flow"]
        direction LR
        n0@{ icon: "mdi:brain", form: "rounded", label: "Google Gemini Chat Model", pos: "b", h: 48 }
        n1@{ icon: "mdi:memory", form: "rounded", label: "Simple Memory", pos: "b", h: 48 }
        n4@{ icon: "mdi:cog", form: "rounded", label: "brave_web_search", pos: "b", h: 48 }
        n5@{ icon: "mdi:cog", form: "rounded", label: "brave_local_search", pos: "b", h: 48 }
        n6@{ icon: "mdi:robot", form: "rounded", label: "Brave Search AI Agent", pos: "b", h: 48 }
        n7@{ icon: "mdi:play-circle", form: "rounded", label: "Brave Search Workflow Start", pos: "b", h: 48 }
        n1 -.-> n6
        n4 -.-> n6
        n5 -.-> n6
        n0 -.-> n6
        n7 --> n6
    end

    subgraph sg1["Brave Search MCP Server Flow"]
        direction LR
        n2@{ icon: "mdi:play-circle", form: "rounded", label: "Brave Search MCP Server Trig..", pos: "b", h: 48 }
        n3@{ icon: "mdi:wrench", form: "rounded", label: "call_brave_search_agent", pos: "b", h: 48 }
        n3 -.-> n2
    end

    %% Styling
    classDef trigger fill:#e8f5e9,stroke:#388e3c,stroke-width:2px
    classDef ai fill:#e3f2fd,stroke:#1976d2,stroke-width:2px
    classDef aiModel fill:#e8eaf6,stroke:#3f51b5,stroke-width:2px
    classDef decision fill:#fff8e1,stroke:#f9a825,stroke-width:2px
    classDef database fill:#fce4ec,stroke:#c2185b,stroke-width:2px
    classDef api fill:#fff3e0,stroke:#e65100,stroke-width:2px
    classDef code fill:#f3e5f5,stroke:#7b1fa2,stroke-width:2px
    classDef disabled stroke-dasharray: 5 5,opacity: 0.5
    class n7,n2 trigger
    class n6 ai
    class n0 aiModel
    class n3 ai
    class n1 ai

The Challenge: Fast research that’s actually reusable

Manual research breaks down in boring, expensive ways. You search, skim, open a few results, and then rewrite the same “background” explanation for the third time this week. Then comes the messy part: someone asks “where did that claim come from?” and you can’t find the original link because it’s buried in your browser history. It’s not just time. It’s the mental load of keeping context straight while switching between search, docs, chat, and whatever tool you’re writing in.

None of these alone is the problem. Together, they are.

You spend about 30–60 minutes just getting to a “good enough” first summary, and it happens again for every new topic.
Search results come back as raw links, so you still have to read and synthesize before you can share anything.
Local-intent questions (“near me,” city names, neighborhoods) take extra effort because you end up running different searches manually.
Everybody summarizes in a different format, which means your briefs feel inconsistent and hard to reuse later.

The Fix: One-question research summaries, delivered on demand

This n8n workflow wraps Brave Search and Google Gemini into a single “research assistant” endpoint you can call from other tools. You send a plain-English question into the workflow (through an MCP trigger or a test start trigger). An AI agent reads the question, decides if it needs a standard web search or a local-style search, then calls the right Brave Search tool through an external MCP service (commonly something like Smithery). Once results come back, Gemini turns them into a clean summary you can actually paste into a brief, a doc, a ticket, or a client update. It also keeps short-term conversational memory during the run, so follow-up questions don’t feel like starting over.

The workflow starts with a single intake message that contains your query. The agent orchestrates Brave web or local lookups, then Gemini produces a structured response. Finally, that summary is returned to the calling client as the output, which means you can plug it into other n8n automations easily.

What Changes: Before vs. After

What This Eliminates

Impact You’ll See

Manually deciding which search approach to use for each question.
Copying links into a doc, then trying to summarize them under time pressure.
Rewriting the same background explanation every time someone asks a similar question.
Juggling separate “search” and “summarize” tools without a repeatable output format.

Most teams get a usable first summary in about 5 minutes instead of an hour.
Fewer “can you clarify this?” messages because the output is consistent.
Less context loss when you ask follow-ups in the same run.
Cleaner handoffs between people, especially for briefs and client notes.
One place to update the prompt, so your voice stays stable over time.

Real-World Impact

Say you produce 5 content briefs a week, and each one needs a quick competitive scan and a “state of the market” paragraph. Manually, it’s easy to burn about 45 minutes per brief across searching, opening results, and writing the summary, so you’re looking at roughly 4 hours weekly. With this workflow, you submit one question, wait a couple of minutes for the agent to pull results and Gemini to summarize, and you’re done. That’s the difference between “research day” and “research slot.”

Requirements

n8n instance (try n8n Cloud free)
Self-hosting option if you prefer (Hostinger works well)
Brave Search (via an MCP provider) to fetch web and local results.
Google Gemini to write the final research summary.
Google AI API key (get it from Google AI Studio)

Skill level: Intermediate. You’ll connect credentials, paste an endpoint URL, and adjust one system prompt.

Need help implementing this? Talk to an automation expert (free 15-minute consultation).

The Workflow Flow

A client sends a query. The workflow can be triggered through an MCP Server Trigger endpoint for real usage, or through the “start trigger” used for testing and direct execution.

Gemini becomes the “brain.” The Google Gemini chat model is connected to the agent so it can interpret what the user actually wants, not just match keywords.

The agent chooses the right search tool. Based on the wording, it routes to Brave web search or Brave local search (both called through MCP client tools provided by your external MCP service).

A clean answer comes back. The workflow returns a single summarized response to the caller, and it can keep short-term context inside the run using a memory buffer.

You can easily modify the system prompt to change formatting, tone, or how citations are handled based on your needs. See the full implementation guide below for customization options.

Step-by-Step Implementation Guide

Step 1: Configure the MCP Server Intake Trigger

This workflow listens for MCP requests and routes them into the Brave search agent pipeline.

Add and open MCP Server Intake.
Set Path to cc8cc827-3e72-4029-8a9d-76519d1c136d.
Confirm Invoke Brave Agent Tool is connected as an AI Tool output from MCP Server Intake (this is how MCP requests are routed to the agent workflow).
Optionally keep Flowpast Branding for in-canvas documentation; it does not affect execution.

Use the exact MCP path value when registering this workflow in your MCP client to ensure requests reach MCP Server Intake.

Step 2: Connect Brave Search Services

The agent relies on Brave MCP tools for web and local searches.

Open Brave Web Lookup and confirm Tool Name is brave_web_search, Operation is executeTool, and Connection Type is http.
Set Tool Parameters to ={{ /*n8n-auto-generated-fromAI-override*/ $fromAI('Tool_Parameters', `Generate a JSON object with the parameters for the 'brave_web_search' tool. The tool expects a JSON object with the following keys: - "query": (string, **required**) The search terms for the web search. - "count": (number, optional) The number of results to return per page. Maximum is 20. If not provided, a default value will be used by the service. - "offset": (number, optional) The pagination offset to retrieve a specific page of results. Maximum is 9. If not provided, the first page is returned. Example (required only): {"query": "latest AI research"} Example (with optionals): {"query": "popular javascript frameworks", "count": 10, "offset": 0}`, 'json') }}.
Open Brave Local Lookup and confirm Tool Name is brave_local_search, Operation is executeTool, and Connection Type is http.
Set Tool Parameters to ={{ /*n8n-auto-generated-fromAI-override*/ $fromAI('Tool_Parameters', `Generate a JSON object with the parameters for the 'brave_local_search' tool. The tool expects a JSON object with the following keys: - "query": (string, **required**) The search terms for the local search (e.g., "coffee shop downtown", "bookstore [zip code]"). - "count": (number, optional) The number of local results to return. Maximum is 20. If not provided, a default value will be used by the service. Example (required only): {"query": "restaurants near Eiffel Tower"} Example (with optional count): {"query": "best parks in Berlin", "count": 5} Note: This tool does not support an 'offset' parameter for pagination. It may fall back to web search if no local results are found.`, 'json') }}.
Credential Required: Connect your mcpClientHttpApi credentials. These tools are connected as AI tools to Brave Search Orchestrator, so ensure credentials are available to the parent agent’s tool connections.

Step 3: Set Up the Brave Search Orchestrator Agent

This agent interprets the user query, chooses the right search tool, and composes the response.

Open Brave Search Orchestrator and set Text to ={{ $('Subflow Start Trigger').item.json.query }}.
Set Prompt Type to define and paste the full System Message from the workflow into the field so the agent understands when to use web versus local search.
Open Gemini Chat Engine and set Model Name to models/gemini-2.5-flash-preview-05-20.
Credential Required: Connect your googlePalmApi credentials. Gemini Chat Engine is connected as the language model for Brave Search Orchestrator, so credentials should be added there.
Open Window Memory Buffer and set Session Key to ={{ $execution.id }} with Session ID Type set to customKey.
Confirm Window Memory Buffer is attached as AI Memory to Brave Search Orchestrator for per-execution context retention.

Step 4: Configure the Subflow Entry and Agent Tool Workflow

The MCP trigger invokes a tool that calls a subflow entry, which then passes the query into the agent.

Open Subflow Start Trigger and ensure the workflow input includes query.
Open Invoke Brave Agent Tool and set Description to =To perform web and local searches using the Brave Search MCP and provide relevant results based on user queries. It should intelligently decide when to use web search versus local search..
In Invoke Brave Agent Tool, map the workflow input query to ={{ /*n8n-auto-generated-fromAI-override*/ $fromAI('query', `The primary natural language query from the user for searching.`, 'string') }}.
Select the target workflow in Workflow ID so it points to the workflow containing Subflow Start Trigger.

⚠️ Common Pitfall: Leaving Workflow ID empty in Invoke Brave Agent Tool will prevent the MCP request from reaching Subflow Start Trigger.

Step 5: Test and Activate Your Workflow

Validate the MCP intake and agent response before production use.

Use Execute Workflow with a sample query input on Subflow Start Trigger to confirm Brave Search Orchestrator returns a summarized response.
Send a real MCP request to the MCP Server Intake path and verify that the agent chooses between Brave Web Lookup and Brave Local Lookup.
A successful run should show tool calls in the execution log and a final response summarizing search results.
Toggle the workflow to Active to accept MCP requests in production.

🔒

Unlock Full Step-by-Step Guide

Get the complete implementation guide + downloadable template

Watch Out For

Google Gemini credentials can expire or need specific permissions. If things break, check your n8n credential entry for the “Google Palm API” (Gemini) connection first.
If you’re using Wait nodes or external rendering, processing times vary. Bump up the wait duration if downstream nodes fail on empty responses.
Default prompts in AI nodes are generic. Add your brand voice early or you’ll be editing outputs forever.

Common Questions

How quickly can I implement this Brave Gemini summaries automation?

About 30 minutes if you already have your API keys and MCP endpoint.

Can non-technical teams implement this Brave Gemini summaries automation?

Yes, but someone needs to be comfortable pasting credentials and testing a webhook-style endpoint. After it’s working once, day-to-day use is just sending questions.

Is n8n free to use for this Brave Gemini summaries workflow?

Yes. n8n has a free self-hosted option and a free trial on n8n Cloud. Cloud plans start at $20/month for higher volume. You’ll also need to factor in Google Gemini API usage and your Brave Search MCP provider costs.

Where can I host n8n to run this automation?

Two options: n8n Cloud (managed, easiest setup) or self-hosting on a VPS. For self-hosting, Hostinger VPS is affordable and handles n8n well. Self-hosting gives you unlimited executions but requires basic server management.

How do I adapt this Brave Gemini summaries solution to my specific challenges?

You’ll mainly adjust the system prompt inside the Brave Search Orchestrator agent to control tone, structure, and what “good” looks like. If your MCP provider uses different tool names or endpoints, swap the credentials on Brave Web Lookup and Brave Local Lookup. Common customizations include adding “include sources,” forcing bullet summaries for briefs, and requiring a short “so what” section for decision-makers.

Why is my Brave Search connection failing in this workflow?

Usually it’s the MCP endpoint URL or header auth being wrong or expired. Re-check the credential used by the Brave web/local MCP client tools, then confirm your provider (for example Smithery) still has your Brave Search tools enabled. If the workflow works in tests but fails in production calls, it can also be a path mismatch on the MCP trigger URL.

What’s the capacity of this Brave Gemini summaries solution?

If you self-host, there’s no execution limit (it’s mostly your server and API rate limits). On n8n Cloud, capacity depends on your plan, but most small teams can run this all day without thinking about it.

Is this Brave Gemini summaries automation better than using Zapier or Make?

Often, yes. This workflow relies on an agent that chooses tools and keeps short-term context, which is awkward to reproduce in “linear” automation builders. n8n also gives you a self-hosting path and more flexible branching without paying per tiny step. Zapier and Make can still be fine if you only need a simple “search once, summarize once” two-action chain and you never plan to reuse the endpoint. If you’re unsure, pick the tool your team will actually maintain, because abandoned automations are just expensive drafts. Talk to an automation expert and we’ll sanity-check the best fit.

Once this is live, research becomes a callable utility instead of a recurring chore. Set it up, tune the prompt, and let the workflow do the repetitive digging.

Brave Search + Google Gemini, instant research summaries

How This Automation Works

n8n Workflow Template: Brave Search + Google Gemini, instant research summaries

The Challenge: Fast research that’s actually reusable