Google Deep Research Max: AI Research Agents Now Support MCP and Custom Data Sources

Google launched Deep Research and Deep Research Max on April 21, 2026, through the Gemini API. Both are available in public preview through paid Gemini API tiers. Google says the agents are built with Gemini 3.1 Pro and are designed for long-horizon research workflows across the open web and custom sources.

The regular Deep Research agent is designed for speed and lower-latency research experiences. Deep Research Max is designed for deeper, more comprehensive work. Google describes Max as suitable for background research jobs, such as generating detailed due diligence reports overnight for analyst teams.

The important update is tool access. Google’s developer documentation says Deep Research supports Google Search, URL Context, Code Execution, MCP Server, and File Search. By default, when no tools are specified, the agent can use Google Search, URL Context, and Code Execution. Developers can restrict or expand the tools depending on the workflow.

MCP support is the most useful part for business research systems. It allows the agent to connect to remote MCP servers, pass authentication headers, and restrict which tools the agent can call. This means a company can connect Deep Research to approved business systems, internal tools, financial data providers, market databases, or domain-specific research sources without forcing everything into a plain web-search workflow.

File Search adds another practical layer. Developers can connect uploaded document corpora and ask the agent to compare private files against public information. Google’s own example shows a fiscal year report being compared against current public web news. This is directly useful for internal knowledge search, board reports, audit preparation, policy research, competitor tracking, and investor research.

Code Execution makes the system more useful for analysis, not only summarisation. The agent can run calculations and data analysis during the research task. For finance teams, this can support market comparisons, revenue trend analysis, valuation summaries, pricing research, and portfolio notes. For education teams, it can help analyse survey data, course outcomes, student progression data, and research papers.

The native chart feature is also important. Google says Deep Research can generate charts, graphs, and other visual elements when visualization is enabled. The documentation recommends explicitly asking for visuals, such as charts showing trends over time or graphics comparing market share. This makes the output more useful for reports, dashboards, strategy documents, and presentation drafts.

Deep Research also supports multimodal inputs, including images, PDFs, audio, and video. Google’s documentation shows examples using image input and document input. This opens up practical workflows such as analysing investor PDFs, research papers, brochures, policy documents, screenshots, recorded material, and course documents as part of a larger research task.

For finance research, the strongest use case is source-heavy analysis. A team can connect public web sources, SEC filings, paid data sources through MCP, internal notes through File Search, and spreadsheet-style calculations through Code Execution. Google also says it is collaborating with FactSet, S&P Global, and PitchBook on MCP server designs for shared customers.

For market analysis, Deep Research Max can be used to track competitors, category movement, pricing changes, funding activity, hiring signals, customer sentiment, and regulatory updates. The value is in producing cited reports with charts and clear sections, rather than a loose summary from a single search session.

For education research, the tool can help teams review policy changes, study labour-market trends, compare programs, analyse public reports, and prepare course-development evidence. An education provider could use it to compare internal course documents with public skill-demand reports, government updates, and competitor program pages.

For internal knowledge search, the practical use case is report generation from mixed sources. A company can search uploaded documents, connected file stores, web sources, and custom MCP tools in one research workflow. That is useful for operations teams that need to prepare monthly reports, compliance notes, market briefs, proposal research, and executive summaries.

The main implementation point is control. Research agents that access private data should have restricted tools, clear permissions, source logging, human review, and defined output formats. Google’s documentation supports collaborative planning, where a user can review and refine the research plan before execution. This is useful when the task is sensitive, expensive, or dependent on specific source coverage.

Google Deep Research Max is not just another research chatbot release. It gives developers a way to build research systems that combine web data, private files, MCP-connected tools, code execution, and generated visuals. For companies already doing manual research across browser tabs, PDFs, spreadsheets, and internal folders, this is a practical path toward research automation.