Our clients have frequently shared how painful it has been to build internal AI agents on top of a fragmented set of MCP solutions. Stitching together a series of disparate data providers, each with different search capabilities and data structures, has left internal tech teams responsible for building and maintaining the intelligence system needed to turn that information into answers and work products with a common set of challenges.
Today we are launching the solution: the SuperAnalyst MCP. SuperAnalyst now works for you both inside AlphaSense and outside, wherever you need work done. It is a true agent MCP you can delegate complex work to, leveraging AlphaSense’s library of more than 500 million business documents, thousands of Canalyst models, and much more through a single endpoint. It plans the research, sources the right information, executes analysis, writes and runs code, and returns the right context for your internal agents to build on with granular source citations to premium content you can trust.
What used to be fragmented and painful becomes seamless and efficient. The full intelligence of our SuperAnalyst agentic system is now available everywhere you need it.
The Cost of Assembling Research From Fragmented Parts
Most teams building research agents today assemble them from parts. Each data vendor ships its own MCP, so a team licenses one for filings, another for transcripts, another for financials, and another for expert calls. That stack of MCPs gets expensive before a single question is asked.
Each of those MCPs has its own search (often not the data vendor's strength), its own schema, and its own idea of what a company or an industry is. None of them shares a context graph with the others, so nothing tells your agent that the supplier named in a transcript is the same company as the one in a filing. Your agent burns tokens fumbling across black boxes and still ends up with blind spots in its research.
Many of our customers started out expecting to plug in dozens of MCPs and let their agents sort it out. They learned quickly that agents can't handle that many. Every MCP adds its tool definitions to the model's context before any work begins. With enough of them, the model picks the wrong tool or skips the right one, calls the same source twice, or loses track of what it has already found — and the errors and blind spots compound.
The harder problem is context. When every MCP stops at retrieval, your harness has to do the research itself. It decides which tools to call and in what order, reads what comes back, reconciles conflicts across sources, and works out when it has enough. Building that harness is hard, and keeping it working as vendors change their APIs is harder. You also can't delegate any of the work. Every step happens inside your model's context window.
That is where the bill grows. An agent works in hops: plan, search, read, reason, search again. On every hop the model is sent the whole conversation so far, including every raw result it has already read. In a typical 10-step task, around 85% of the token bill goes to context the model has already seen. Every extra search against an extra MCP adds to that pile, and the cost compounds on every hop that follows. Quality suffers too. Long, diluted context degrades answers while it inflates cost.
To keep the bill in check, many teams are forced to cap how much individual users can leverage the best models. Moving to cheaper models saves on per-token cost, but every answer is now written by a less capable model, at exactly the point where the research needs the most judgment.
One MCP Built as a Specialized Research Agent
We built the SuperAnalyst MCP so teams can stop assembling and start delegating.
- One connection to the sources that matter. Expert calls, channel checks, earnings transcripts, filings, news, models, financials, and the rest of the AlphaSense library sit behind a single MCP, under one license.
- Everything is tagged and enriched the same way. Every document carries the same tickers, companies, industries, and entities. It is linked into one context graph of supply chain relationships and key debates, so your agent can follow relationships across sources that a vector search would miss.
- A harness that is already built and tuned. Query planning, multi-step research, source weighting, and context management are built in and tuned on financial research, across every source at once.
- One standard of relevance. The system judges credibility, freshness, and authority seamlessly across a conference transcript, an expert call, or a 10-K, while understanding how these perspectives relate to one another so your agent doesn't have to reconcile different ranking schemes.
- Citations on every claim. Every piece of evidence links seamlessly back to the passage in the source document, so whatever your model does next stays traceable and works the same way across every source.
- An agent you can delegate to. Give SuperAnalyst an objective, not a sequence of tool calls. It works out the approach, draws on the skills the task needs, and returns the best response for the prompt: that might be a direct answer, an Excel workbook, or a cited evidence pack.
The work splits cleanly. Your agent leads: it holds your personal context and decides what it needs from AlphaSense. SuperAnalyst is the research agent beneath it. It assembles evidence in fewer turns than a general model searching on its own, so more of your model's budget goes to the reasoning that decides the quality of the work.
What That Changes, in Numbers
We test this setup the same way we test every new model: by cost per correct answer, counting every token in the research loop and not only the list price.
A frontier model doing its own retrieval over our content with raw vector search, a good proxy for a typical market data MCP, costs roughly 3x more per question than the same model running in our harness with AlphaSense Search. The AlphaSense answers are also preferred more than 2:1 even when we give the alternative our proprietary content.
Those gains come from the system around the model: content, search, pre-computed context, tools, and harness. So they carry over to every new model. When a better model arrives, you plug it in above the SuperAnalyst MCP and keep the research advantage underneath it. Your skills, workflows, and integrations stay as they are. You never have to choose between the best model and an affordable bill.
What You Can Build With It
When research is something your agent delegates rather than something it does itself, new use cases become practical.
A red team agent. Hand it a thesis, a pitch, or an investment memo and ask it to argue the other side. SuperAnalyst finds the strongest counter-evidence across expert calls and management commentary. It shows where the bear case is gaining ground and which assumptions the evidence doesn't support, with every point cited, so your team can stress-test a view before it reaches the investment committee.
A supply chain risk agent. Give it a company or a portfolio and ask where the supply chain risk sits. SuperAnalyst starts from our supply chain knowledge graph, so it skips the mapping and spends its budget going deeper: suppliers' suppliers, shared single-source dependencies, and the filings and expert calls behind each exposure, all cited.
A meeting-prep agent. Your agent reads the deal history and account notes in your CRM, then asks SuperAnalyst for the market view with that context included. The brief reflects both your relationship and the latest on the company.
An always-on coverage monitor. An agent that watches a coverage list or a portfolio and asks SuperAnalyst each morning what changed overnight: guidance, management changes, expert sentiment, channel checks.
An evidence pack for downstream workflows. A model update or memo-drafting pipeline asks for exactly the evidence it needs and gets back a focused, cited set of expert call quotes and management commentary, ready to feed the next step.
Bringing In Your Own Content
Your firm's research, models, meeting notes, and house views hold years of judgment that no one outside your walls has seen. That knowledge is a core part of your competitive advantage, and it is most powerful when it meets the best market intelligence. When SuperAnalyst reasons across both, it can test your house view against our content, connect your internal notes to what is moving in the market, and produce insight that reflects how your firm thinks. And with our SuperAnalyst MCP, we give you flexibility to bring that intelligence together in a way that works for every use case.
- Index it with AlphaSense (recommended). Thousands of clients already have AlphaSense securely index their internal content alongside our library. Your documents get the same tagging enrichment and relevance ranking as everything else, so SuperAnalyst can reason across your content and market intelligence in a single request.
- Pass it in through the prompt. Send context to SuperAnalyst on the fly through the MCP, for example, a draft memo, a set of meeting notes, or a model. You’re in control of security and retention policies.
- Combine it on your side. Keep your content in your own systems and bring the SuperAnalyst MCP into your agents and applications there. Your orchestration decides when a task calls for SuperAnalyst and merges its cited output with your internal sources.
SuperAnalyst, Everywhere
SuperAnalyst is the same agent wherever you use it. In the AlphaSense platform, it works alongside your team. Through the SuperAnalyst MCP, it works inside the AI assistants your team already uses and the applications you build. The content, context graph, harness, and citations are the same either way.
Now that models can delegate, an MCP should do more than retrieve. It should take on the research, decide what matters, and return what the task needs, at a cost that lets your agents keep working. That's the standard we built the SuperAnalyst MCP to meet.
