MCP servers are what let AI assistants like Claude and ChatGPT pull live data from the tools your firm already uses, and the number of options has grown faster than anyone's ability to evaluate them.
This article is a curated, opinionated take on which MCP servers actually pay off for a private capital firm: which to prioritize, what each one unlocks, what each one can’t do, and the recommended starter stack for a 30-person PE buyout shop and a 10-person VC. It's published by us at Affinity, and it's honest about where Affinity is the answer and where it isn't.
What an MCP server actually does, and why ops leads should care
An MCP server lets an AI assistant (Claude, ChatGPT, Gemini) query a specific data source and return structured answers instead of guessing or browsing the web. Without an MCP, asking Claude "what's the latest valuation on Stripe?" returns a guess based on training data that's months old. With the PitchBook MCP connected, it returns the actual current data.
Every MCP replaces a multi-step workflow: open the tool, search, copy, paste into the chat. With the MCP connected, the AI does that retrieval in the background. The value depends entirely on which MCPs you connect. With the wrong stack you've created complexity. With the right stack, you've shortened ten workflows your team runs every week.
How to evaluate any MCP server: the three-question test
Each MCP section that follows answers three questions. They're the same questions an ops lead should ask before wiring up anything:
What does it actually unlock? Specific workflows the AI can now run that it couldn't before. If the answer is vague, the MCP probably isn't ready.
What's the setup cost? Some MCPs are a five-minute OAuth flow. Others require an existing enterprise deployment and IT involvement. The difference matters when you're trying to show value before Friday's partner meeting.
What doesn't it do? Every MCP has a boundary. Knowing that boundary before you wire it up saves weeks of mismatched expectations.
The bundled option: Anthropic's financial-services plugin marketplace
Anthropic ships a financial-services plugin marketplace that bundles connectors for PitchBook, Capital IQ, Morningstar, Moody's, Chronograph, FactSet, Dun & Bradstreet, and Egnyte. One install, eight connectors. For firms starting from zero on MCP and wanting one decision instead of eight, this is the fastest path to coverage. PE buyout shops with LP reporting obligations get Chronograph in the bundle, which is the right answer for them anyway. Firms whose AI workflows include public-markets and credit analysis alongside private capital get breadth they'd otherwise assemble piecemeal.
The trade-off: the bundle doesn't include the relationship and deal history layer or VC-specific sourcing tools. Smaller VC firms that won't use most of what's in the bundle may prefer to install individual MCPs. For a deeper walkthrough of what's in the bundle, see the companion piece on Anthropic's finance agent templates and what they need to work for PE and VC.
The financial data layer: PitchBook, Capital IQ, Morningstar
PitchBook's MCP server connects Claude to profiles on more than 8.5 million private companies, 2.8 million deals, 597,000 investors, and 154,000 funds. For a deal team, this means the AI can pull live company profiles, cap tables, recent funding rounds, comparable transactions, and fund performance data without leaving the chat. Setup requires an existing PitchBook license and SSO credentials. The connection is cloud-managed, so there's no local installation or custom integration. A firm with an existing PitchBook subscription gets the highest immediate payoff from wiring this up first.
Capital IQ and Morningstar cover adjacent territory. Capital IQ is stronger on public-company financials and credit data while Morningstar adds independent analyst-backed intelligence across both public and private markets. All three ship inside the Anthropic bundle for firms that want them together.
However, none of these tell the AI what your firm knows about a company. They cover who a company is, but not your relationship history, pipeline context, or prior interactions. That's a different layer.
The relationship and deal history layer
When a partner asks Claude "who at our firm knows the CEO of [target]?", the financial data MCPs have nothing to offer. Without the relationship layer, the AI either refuses, fabricates, or suggests a LinkedIn search. With Affinity's MCP server, the answer draws from the firm's actual data: who emailed whom, when the last meeting happened, what notes were logged after the call, and how strong the relationship is across the team.
The same applies to "what's the last note on this deal?", "draft a pre-meeting brief using our actual history with this company", and "which companies in our pipeline have a connection to this LP?" These are questions that require the firm's own data, and no third-party provider has that data because it's generated by the firm's own communications and workflows.
Setup cost depends on starting point. For firms already on Affinity, the MCP server is a single connection in admin settings. For firms on a different CRM or on spreadsheets, the conversation is about CRM infrastructure first and MCP second.
Here’s what the Affinity MCP doesn’t do: replace PitchBook for valuation lookups, pull portfolio company financials, or run financial models. It's the layer that gives the AI context about your firm's relationships and deal history. Every other MCP becomes more useful when this layer is connected, because the AI can prioritize results based on who and what actually matters to your fund.
Already have Affinity's MCP server connected? Find a starter set of AI workflows here.
The portfolio data layer: Chronograph and Standard Metrics
Two tools serve this layer, built for different firm profiles.
Chronograph is the institutional answer, serving clients managing more than $4 trillion in private capital. The MCP connects AI tools to fund performance data, company metrics, exposure analytics, and LP reporting workflows. In March 2026, Chronograph added semantic search, bridging unstructured documents with structured portfolio data so the AI can query both through a single interface. Setup requires an existing Chronograph deployment.
Standard Metrics serves VC firms that need portfolio monitoring without institutional weight. More than 100 VC firms and 10,000 companies run on Standard Metrics. The MCP connects Claude to revenue, growth rate, burn rate, and runway data. Setup is an OAuth flow against an existing subscription.
Are these MCPs useful for your team? PE buyout shops and growth equity firms with serious LP reporting obligations should go with Chronograph. Smaller VCs with simpler portfolio reporting should go with Standard Metrics. We think running both would be unusual.
The VC-specific sourcing layer: Harmonic
While PitchBook covers companies once they reach a certain threshold of visibility, Harmonic covers what happens before that.
Harmonic's MCP connects Claude to talent movement data, company formation activity, hiring patterns, and pre-formation signals. A principal researcher leaves a foundation model lab. Two engineers with complementary backgrounds stop updating LinkedIn. A repeat founder goes quiet after an acquisition. These are the patterns that VCs running serious top-of-funnel work track, and Harmonic's data captures them at a stage where PitchBook has nothing to show.
Setup requires an existing Harmonic subscription and OAuth credentials. The MCP exposes 13 tools for company search, enrichment, and similarity matching.
A note on stack design: Harmonic integrates natively with Affinity, so data flows between the sourcing layer and the relationship layer without a custom integration. For VC firms running both, the AI gets a more complete picture of any single company. Harmonic tells it what the company looks like from the outside and Affinity tells it what the firm already knows from the inside.
Skip Harmonic if you're a PE buyout shop focused on mature companies. This is built for the pre-PitchBook stage of a company's life.
What's notably absent from the MCP landscape
Three gaps worth watching:
Expert call data
AlphaSense (which acquired Tegus in 2024) has published a reference implementation of an MCP server in its developer documentation. It exposes one tool, AlphaSense Search, for AI-powered research with citations. It's a developer sample meant to run locally, not a hosted enterprise product. Firms whose workflows depend on expert call libraries are still working out of AlphaSense directly. This is worth tracking as it matures.
Private market sourcing intelligence
Grata and Sourcescrub are both now part of Datasite, which acquired Valu8 in May 2026 to expand its private market intelligence coverage. Datasite launched an MCP server in late April 2026, but it's focused on virtual data room content (Datasite Prepare and Diligence), and not Grata's company sourcing data. For firms whose sourcing workflow runs on Grata, the most consequential MCP in the landscape hasn't shipped yet.
Both Grata and Sourcescrub integrate natively with Affinity.
Bloomberg and Refinitiv
Community-built MCPs exist on GitHub, but no official enterprise-grade MCP has been introduced from either provider. For most private capital firms, PitchBook and Capital IQ already cover this category.
Of course, the MCP landscape moves quickly. A firm setting up its stack today should expect to revisit in six months or less.
The recommended starter stack for a 30-person PE buyout shop
Three MCPs cover the core workflow surface area from sourcing through post-close.
Start with Affinity if the firm is already a customer. The AI immediately starts answering questions about the firm's own relationships and pipeline. Add PitchBook second (or the Anthropic bundle for Capital IQ and Morningstar alongside it), and Chronograph third. It requires a longer setup, but it's the right institutional answer for portfolio data and LP reporting.
Skip Harmonic (pre-PitchBook-stage companies aren't where buyout sourcing lives) and Standard Metrics (underbuilt for PE-scale reporting), and ignore generic developer-list MCPs like Filesystem, Puppeteer, and Brave Search. They're built for engineers, not investors.
The recommended starter stack for a 10-person VC firm
This team has different priorities and requires a different stack.
Again, start with Affinity for the same reason. The firm-specific context is the layer that makes every other MCP more useful. Add PitchBook second for company data, and Harmonic third, because top-of-funnel sourcing is where most VCs get the highest firm-specific payoff. Standard Metrics should be fourth if the portfolio warrants automated monitoring.
Skip Chronograph unless LP reporting is a serious obligation, as it's overbuilt for 10-person VCs. If you'd rather install once, the Anthropic bundle gives you PitchBook plus credit and public-markets connectors at no extra setup cost.
What to do this quarter
Audit existing subscriptions. The MCP for a tool you already pay for is the highest-ROI install. Anything else is a budget conversation before it's a technology conversation.
Wire up the relationship and deal history layer first. It's the layer that makes every other MCP more useful, because it tells the AI which companies and people actually matter to your firm. If the firm is already on a CRM with an MCP server, setup is measured in minutes.
Hold the line on developer-list MCPs. They are designed for software engineers, not investors. Adding them creates noise that often makes the AI less useful.
Frequently asked questions
Does PitchBook have an MCP server?
Yes. PitchBook's Premium Connector is built on MCP server technology and connects to Claude, giving deal teams access to profiles on more than 8.5 million private companies, 2.8 million deals, and 154,000 funds directly in the AI assistant. Setup requires an existing PitchBook license and SSO credentials.
Does Chronograph have an MCP server?
Yes. Chronograph launched its MCP integration in partnership with Anthropic, connecting private capital portfolio data (fund performance, company metrics, exposure analytics, LP reporting) to Claude. It's available to existing Chronograph clients with a paid Claude subscription.
What does the Affinity MCP server do?
Affinity's MCP server connects a firm's relationship graph, deal pipeline, and interaction history to AI assistants including Claude, ChatGPT, and Gemini. It enables queries like "who at our firm knows this CEO?", "what's our last interaction with this company?", and "draft a brief using our relationship history." It's the layer that gives AI agents firm-specific context that no third-party data provider has.
What does the Harmonic MCP server do?
Harmonic's MCP server connects AI assistants to talent movement data, company formation activity, and pre-formation signals. It's designed for VC firms doing top-of-funnel sourcing, tracking companies before they reach the visibility threshold where PitchBook and similar platforms begin coverage. It integrates natively with Affinity.
What is the difference between Chronograph and Standard Metrics?
Chronograph is institutional-grade portfolio monitoring for PE firms, growth equity firms, and LPs, serving clients managing more than $4 trillion in private capital. Standard Metrics is lighter-weight portfolio monitoring built for VC firms, serving 100+ VC firms and 10,000+ companies. PE firms with complex LP reporting want Chronograph. Smaller VCs with simpler reporting want Standard Metrics.
How long does it take to set up an MCP server?
Setup time varies. Cloud-managed MCPs like PitchBook's require only SSO authentication and can be connected in minutes. CRM-based MCPs like Affinity's are similarly fast for existing customers. Institutional platforms like Chronograph require an existing enterprise deployment and longer setup timelines.
Are MCP servers secure for investor data?
MCP servers enforce the same permissions and access controls as the underlying platform. PitchBook data respects license entitlements. Chronograph enforces client-level data separation. Affinity's MCP server inherits the firm's existing admin settings. The AI only accesses data the user is already authorized to see.
Which MCP servers should a small VC firm use?
A 10-person VC firm should start with Affinity (relationship and deal history), add PitchBook (company data), then Harmonic (pre-formation sourcing intelligence). Standard Metrics is the fourth add if the portfolio is large enough to warrant automated monitoring. Skip Chronograph unless LP reporting is a major obligation.
Which MCP servers should a PE buyout firm use?
A 30-person PE buyout firm should start with Affinity (relationship and deal history), add PitchBook or the Anthropic financial-services bundle (company and market data), then Chronograph (portfolio monitoring and LP reporting). Skip Harmonic and Standard Metrics, which are built for VC workflows.
What MCP servers are not yet available for private capital workflows?
As of May 2026, the most notable gaps are: Grata/Sourcescrub sourcing data (Datasite owns both but the sourcing MCP hasn't shipped; only the VDR MCP has), AlphaSense expert call data (a developer reference implementation exists but no hosted enterprise product), and Bloomberg/Refinitiv (community MCPs exist but no official enterprise-grade servers).
For the deep-dive on Affinity's MCP and how it connects your firm's relationship and deal history to Claude: Affinity MCP server.
.png){kind=logo}
.webp)
.webp)
