The Best MCP Servers for Finance in 2026: Banking, Lending, Credit & Financial Data
- Jason Feimster
- 15 hours ago
- 16 min read
Compare the best MCP servers for finance in 2026, including Ramp, Mercury, Plaid, QuickBooks, LoanPro, D&B, Moody’s, Stripe and more. See which servers fit banking, lending, business credit, accounting, payments, treasury and financial-data workflows—and where human controls still matter.
AI agents are finally getting access to the systems where money actually lives.
The best MCP servers for finance in 2026 include Ramp for finance operations, Mercury for business banking, Plaid for financial-data infrastructure, QuickBooks for accounting, LoanPro for lending, Dun & Bradstreet and Moody’s for commercial credit and risk, Modern Treasury for money movement, Stripe and PayPal for payments, and Alpha Vantage or Morningstar for market and investment data.
That sounds like a software shopping list.
It is not.
The important question is which financial system you want an AI agent to read, analyze or act on—and how much authority you are willing to give it.
A server that reads a bank balance and a server that can create a payment order both speak MCP.
Those are not remotely the same risk.
This guide compares the strongest finance MCP servers available in 2026 by actual financial function rather than tossing 47 GitHub projects into a table and declaring victory.
Best Finance MCP Servers at a Glance
MCP Server | Best For | Finance Category | Access Model | Our Take |
|---|---|---|---|---|
Ramp MCP | Business finance operations | Spend / Finance Ops | Authenticated remote | Best overall for operating finance |
Mercury MCP | Live business banking data | Banking | OAuth, read-only | Best banking MCP |
Plaid MCP | Fintech integrations and diagnostics | Open Banking / Data | Remote + local tooling | Best for fintech builders |
Stripe MCP | Payments and revenue workflows | Payments | Remote OAuth / local | Best payment infrastructure MCP |
PayPal MCP | Merchant payments and invoices | Payments / Commerce | Remote + local | Best merchant-focused option |
QuickBooks Online MCP | Accounting and financial reports | Accounting | Local OAuth | Best accounting MCP |
Modern Treasury MCP | Payment operations and ledgers | Treasury | Local package/API credentials | Best for money-movement infrastructure |
Brex MCP | Spend, cards and expenses | Finance Ops | Authenticated | Best Ramp alternative |
LoanPro MCP | Loan servicing and lending workflows | Lending | Enterprise platform access | Best lending-specific MCP |
Dun & Bradstreet MCP | Business credit and company risk | Commercial Credit | Enterprise MCP access | Best business-credit data |
Moody’s MCP | Credit analysis and risk intelligence | Credit / Risk | Remote enterprise | Best institutional credit intelligence |
Alpha Vantage MCP | Stocks, FX and market data | Financial Data | Remote/API key | Best accessible market-data MCP |
Morningstar MCP | Investment research and portfolios | Investment Data | Licensed enterprise access | Best institutional investment research |
There is no universal winner.
The best finance MCP server is the one attached to the financial system that already owns the data or workflow you need.
That distinction matters more than tool count.
What Is an MCP Server for Finance?
Model Context Protocol, or MCP, is an open standard that gives AI applications a consistent way to connect with external tools, data sources and services.
An MCP server can expose tools, resources and prompts to an AI application. In practical terms, that means an assistant can retrieve information or perform approved actions against a live system instead of relying entirely on what the model remembers. Model Context Protocol specification
For finance, that can mean:
reading bank balances;
analyzing transactions;
retrieving invoices;
reviewing a P&L;
inspecting payment activity;
accessing commercial credit data;
researching a borrower's business;
reviewing a loan account;
analyzing portfolio holdings;
generating finance reports;
or, with sufficient permissions, initiating actual financial actions.
That last line is where the adult supervision enters the room.
MCP does not magically make an AI model a banker, accountant, underwriter or CFO.
It gives the model structured access to the tools those professionals use.
That can be enormously useful.
It can also be an excellent way to automate a bad decision at impressive speed.
1. Ramp MCP — Best Overall MCP for Finance Operations
Best for: CFO teams, founders, finance operators and businesses already using Ramp.
Ramp has moved unusually aggressively into MCP. The company now describes three separate MCP offerings:
Ramp MCP for interacting with Ramp account data and workflows;
Developer MCP for developers building against Ramp APIs;
Ramp Data MCP for querying public benchmark datasets such as Ramp Rate and its AI Index.
The main Ramp MCP can support workflows around transactions, expenses, approvals and spend analysis. Ramp states that a user's existing role and permissions carry through to the connection, with writes recorded in the Ramp Audit Log. Ramp MCP documentation · Ramp support
Why it stands out
Ramp is not just exposing documentation.
It is turning the finance operating system itself into an agent-accessible layer.
That creates practical prompts such as:
What did we spend on software last quarter?
Which transactions are missing receipts?
Show me expenses outside policy.
Where did travel spend increase this month?
For SMB finance operations, this is closer to an AI finance operator than a generic API wrapper.
Verdict: Best overall if Ramp is already part of your financial stack.
2. Mercury MCP — Best MCP Server for Business Banking
Best for: Founders and businesses wanting AI access to live banking information without giving the model permission to move money.
Mercury's hosted MCP server connects compatible AI applications directly to Mercury account data through OAuth.
The interesting decision Mercury made is architectural:
its hosted MCP is read-only.
Users can query balances, transactions, statements, cards, invoices, credit accounts, customers and recipient information, but Mercury intentionally prevents the MCP server itself from initiating transactions. Mercury MCP documentation
Mercury explicitly positions that limitation as a protection against unintended financial actions. Mercury security best practices
Why it stands out
For financial AI, sometimes the missing feature is the feature.
A useful agent could answer:
What is our current cash position?
How has spending changed over six months?
Show our outstanding invoices.
Which account had this transaction?
How much available credit do we have?
without also possessing a glowing red SEND MONEY button.
Verdict: One of the best-designed banking MCPs for teams that want analysis before autonomy.
3. Plaid MCP — Best for Fintech Builders and Open-Banking Infrastructure
Best for: Developers and fintech teams already building with Plaid.
Plaid currently offers two MCP approaches:
a hosted Dashboard MCP server for production diagnostics and analytics; and
a local MCP server inside its AI coding toolkit for development tasks such as documentation search, mock data, sandbox tokens and webhook simulation.
There is an important nuance here.
Plaid's Dashboard MCP should not be confused with “give ChatGPT unrestricted access to everybody's bank transactions.”
Its current official use case is primarily helping teams operate and troubleshoot Plaid integrations—including Link conversion, Item diagnostics, usage and development workflows. Plaid product overview
Why it matters for lending
Plaid sits close to some of the most valuable data flows in modern underwriting: connected financial accounts, transaction data and cash-flow infrastructure.
That makes its MCP strategy particularly interesting for lenders, brokers and fintech builders even when the official MCP itself is currently focused on operating the Plaid integration rather than functioning as a generic consumer-finance feed.
Verdict: Best MCP for teams building financial-data products rather than merely analyzing their own business account.
4. Stripe MCP — Best MCP for Payments and Revenue Infrastructure
Best for: Developers, SaaS companies and platforms already built around Stripe.
Stripe's official MCP server allows AI agents to interact with Stripe APIs and search Stripe documentation.
Its exposed tools include account information, balances and numerous payment-related resources and actions. Stripe supports a hosted MCP endpoint with OAuth as well as local configurations. The product remains labeled a public preview. Stripe MCP documentation
Where Stripe MCP gets interesting
Imagine an operator asking:
What is our current Stripe balance?
Find the customer associated with this payment.
Show recent subscription activity.
Help me troubleshoot this integration.
Create this approved billing object.
Now compare that with exporting a CSV, opening three browser tabs and asking somebody from engineering what the API means.
That is the shift MCP enables.
The warning
Stripe itself recommends human confirmation around tools and specifically warns about prompt-injection risk when agents interact with multiple servers. Stripe MCP security guidance
When the software touches real money, “the agent seemed confident” is not a control framework.
Verdict: The strongest general payments MCP for businesses whose revenue stack already runs through Stripe.
5. PayPal MCP — Best for Merchant Payments and Invoicing
Best for: Merchants and developers operating inside PayPal.
PayPal offers both local and remotely hosted MCP options and operates official sandbox and production endpoints.
Its server enables merchants to use natural language for business tasks such as retrieving invoice information and working across PayPal capabilities. PayPal expanded the remote MCP server to cover the broader toolset from its agent toolkit. PayPal MCP server
PayPal also documents integration with both Anthropic and OpenAI models. PayPal LLM integration guide
PayPal vs Stripe MCP
The choice is less philosophical than people make it.
If your transaction infrastructure is Stripe, use Stripe.
If your merchant workflow lives in PayPal, use PayPal.
MCP is useful precisely because you should not need to rebuild your finance stack around whichever vendor published the sexiest AI demo this week.
Verdict: Excellent for merchants already embedded in the PayPal ecosystem.
6. QuickBooks Online MCP — Best MCP for Accounting
Best for: Small businesses, accountants, developers and finance teams using QuickBooks Online.
Intuit maintains an official open-source QuickBooks Online MCP server.
The repository currently describes support for 145 tools, 29 QuickBooks entity types and 11 financial reports, including the P&L, balance sheet and cash-flow statement. It runs locally and uses OAuth to authenticate with a QuickBooks Online company. Intuit QuickBooks Online MCP repository
What this changes
Instead of asking:
Can somebody export last quarter's P&L and send it to me?
an appropriately connected assistant can work with the actual accounting system.
That creates useful applications around:
accounts receivable;
vendor activity;
customer balances;
cash flow;
financial reporting;
transaction investigation;
bookkeeping workflows.
For many SMBs, the accounting MCP may actually be more useful than the banking MCP because the books contain the financial classification and context that raw transactions do not.
Verdict: The obvious starting point for QuickBooks-heavy small businesses building an AI finance layer.
7. Modern Treasury MCP — Best for Treasury and Payment Operations
Best for: Fintech companies and finance teams dealing with ledgers, counterparties and serious money movement.
Modern Treasury introduced official MCP support through its TypeScript SDK and publishes its MCP server as an NPM package.
The server translates Modern Treasury API functionality into MCP tools, including functionality around internal accounts, counterparties, expected payments and payment orders. Modern Treasury MCP announcement
This is a different animal from a reporting connector.
Modern Treasury sits in the actual payment-operations plumbing.
Translation
Mercury MCP might answer:
How much cash do we have?
Modern Treasury can live closer to:
Prepare the operational objects required for this payment workflow.
That is powerful.
It also means permissions and human approvals matter dramatically more.
Verdict: Best fit for sophisticated fintech and treasury infrastructure rather than someone who just wants ChatGPT to categorize lunch.
8. Brex MCP — Best Alternative for Spend and Corporate Cards
Best for: Startups and finance teams already operating on Brex.
Brex MCP provides conversational access to financial information including spending, cards, expenses, transactions and reimbursements.
Brex says its MCP respects existing role-based permissions and supports both read and write functionality across core Brex resources. The offering is currently marked early access. Brex MCP overview
Brex vs Ramp
Both are increasingly becoming AI-accessible finance operating systems.
The correct decision is usually the boring one:
Use the MCP that connects to the spend platform you already use.
Migrating corporate finance infrastructure because an MCP comparison gave one tool a gold medal would be spectacularly missing the point.
Verdict: Strong choice for Brex-native finance teams.
MCP Servers for Lending and Credit
This is where most generic finance MCP lists become suspiciously quiet.
Payments are easy to demonstrate.
Credit is harder.
Credit decisions involve identity, financial data, policy, risk models, compliance, documentation, audit trails and judgment.
That is also exactly why MCP could become unusually important in lending.
9. LoanPro MCP — Best Lending-Specific MCP Server
Best for: Lenders, loan servicers and fintechs building AI into lending operations.
LoanPro has built MCP functionality specifically around lending and servicing workflows.
The company positions its MCP layer as a model-agnostic gateway between AI systems and its lending platform, with programmatic compliance controls and auditing. Potential workflows include servicing, collections, underwriting support and answering questions against loan-account data. LoanPro MCP overview
Why this one matters
A generic finance AI knows finance language.
A lending MCP can understand the actual system of record for a loan.
That creates a very different workflow:
inspect loan status;
review payment history;
surface borrower context;
identify servicing conditions;
assemble explanations;
recommend approved next actions;
route consequential actions through controls.
For lending institutions, this is closer to where AI becomes infrastructure rather than another chatbot floating beside the CRM.
Verdict: The most strategically important MCP on this list for lenders and loan-servicing operators.
10. Dun & Bradstreet MCP — Best for Business Credit and Commercial Risk
Best for: Commercial lending, supplier risk, business verification, credit teams and B2B underwriting.
Dun & Bradstreet launched MCP access to its Commercial Graph in 2026.
The MCP offering gives AI systems access to verified business identity, ownership, relationship, payment, financial-strength, credit and risk information. D&B specifically describes finance workflows that include commercial credit origination and portfolio management. Dun & Bradstreet announcement
This matters enormously for business lending.
A business is not merely:
LLC name + revenue + owner's FICO.
Commercial underwriting can depend on questions such as:
Is this the correct legal entity?
Who owns it?
What companies are related?
What does its payment history look like?
What business-risk signals exist?
Is the applicant's stated business consistent with verified records?
D&B's MCP turns that commercial identity layer into something an AI workflow can query directly.
Moonshine operator angle
For commercial brokers and funding platforms, this is the more interesting frontier than asking a generic chatbot whether a company “looks legit.”
Give the agent better data.
Then let the agent reason.
Verdict: Best MCP server for business credit and commercial-company intelligence.
11. Moody’s MCP — Best for Institutional Credit and Risk Intelligence
Best for: Banks, institutional lenders, credit analysts, risk teams, M&A and enterprise finance.
Moody's MCP server exposes proprietary credit ratings, research, entity intelligence and analytical frameworks for risk assessment.
Moody's specifically positions MCP for workflows involving credit analysis, compliance, M&A evaluation and business development. Moody's MCP overview
Its GenAI-ready data spans hundreds of millions of public and private entities, and Moody's has also deployed native MCP integrations with platforms such as Claude. Moody's Claude integration
D&B vs Moody's
There is overlap, but the center of gravity differs.
D&B: commercial business identity, trade/payment insight and business credit.
Moody's: institutional-grade credit ratings, risk research and analytical intelligence.
For a small-business funding agency, D&B may be the more immediately relevant layer.
For a bank credit team or institutional analyst, Moody's can be substantially deeper.
Verdict: Best enterprise MCP for serious credit and risk analysis.
MCP Servers for Financial and Market Data
12. Alpha Vantage MCP — Best Accessible Financial Market Data Server
Best for: Developers, analysts, research agents and market-data workflows.
Alpha Vantage operates an official MCP server that exposes its financial-market functions as MCP tools.
That includes real-time and historical market data, company information, technical indicators and other market datasets. It offers a hosted MCP connection and works with major MCP-compatible agent platforms. Alpha Vantage MCP
Why it makes the list
It provides one of the cleaner routes from:
I need market data in this agent
to:
My agent can actually call market-data functions.
That makes it useful for research agents, investment dashboards and financial-analysis applications without requiring an enterprise data contract to experiment.
Verdict: Best accessible starting point for market-data MCP workflows.
13. Morningstar MCP — Best for Investment Research and Portfolio Intelligence
Best for: Wealth managers, investment teams, portfolio analysts and institutional research.
Morningstar's MCP server brings its investment data and independent research directly into AI workflows.
The platform supports use cases including investment research, portfolio monitoring, analysis and automated reporting while emphasizing traceable, governed data. Morningstar MCP server
This is not the “give me today's AAPL price” layer.
This is closer to:
Ground this investment research workflow in licensed Morningstar intelligence.
Verdict: Best fit when trusted investment research matters more than inexpensive raw market data.
Best MCP Servers for Finance by Use Case
Best for business banking
Mercury MCP — Read-only financial context makes it unusually sensible for AI banking analysis.
Best for finance operations
Ramp MCP — Best combination of spend data, workflow access and operational finance functionality.
Best for accounting
QuickBooks Online MCP — Use the accounting system as the financial source of truth instead of repeatedly exporting reports.
Best for fintech/open banking development
Plaid MCP — Excellent for operating, debugging and building Plaid-powered financial-data infrastructure.
Best for payments
Stripe MCP — Especially strong for SaaS, marketplaces and API-first companies.
Best for merchant commerce
PayPal MCP — Strong for businesses already operating PayPal payment and invoicing workflows.
Best for treasury infrastructure
Modern Treasury MCP — Built much closer to ledgers and money-movement operations.
Best for lending
LoanPro MCP — The most directly lending-native MCP in this comparison.
Best for business credit
Dun & Bradstreet MCP — Verified business identity, payment and commercial-risk data are unusually valuable for B2B credit workflows.
Best for institutional credit analysis
Moody’s MCP — Decision-grade ratings, research and risk intelligence.
Best for market data
Alpha Vantage MCP — Accessible and developer-friendly.
Best for investment research
Morningstar MCP — Best when licensed research, governance and portfolio intelligence matter.
How to Choose an MCP Server for Finance
Do not start by asking:
Which MCP server has the most tools?
Start with five much less exciting questions.
1. What is the system of record?
If your books live in QuickBooks, begin there.
If banking is Mercury, connect Mercury.
If spend is Ramp, connect Ramp.
If loan servicing runs on LoanPro, the lending system matters more than some generic “finance MCP.”
MCP is supposed to remove integration friction.
Do not respond by creating more architecture than the problem requires.
2. Does the agent need read access or write access?
Reading a balance is one risk class.
Issuing a refund, creating a payment order or modifying accounting data is another.
Start read-only whenever the workflow permits it.
3. Who controls authentication and permissions?
Prefer:
vendor-hosted MCP servers;
official vendor implementations;
OAuth or tightly scoped API credentials;
role-based permissions;
audit logs;
explicit write confirmation.
A mystery GitHub repository with four stars and a README that says “just paste your production banking token here” is not fintech innovation.
It is a cry for help.
4. Is the underlying data appropriate for the decision?
Market data is not business credit data.
Bank transactions are not accounting records.
Accounting records are not lender policy.
Business credit reports are not loan approvals.
MCP connects systems. It does not erase the differences between those systems.
5. What requires human approval?
The official MCP specification recommends human oversight for sensitive tool invocations and emphasizes input validation, access controls, rate limits and auditability. MCP tools specification
In finance, that should not be treated as optional decoration.
The Finance MCP Stack I Would Build First
Most businesses do not need thirteen MCP servers.
They need a small stack with clear jobs.
Small-business finance stack
Mercury or banking MCP — live cash and transaction context
QuickBooks MCP — categorized accounting and financial statements
Ramp or Brex MCP — spend, cards and expense operations
Stripe or PayPal MCP — revenue and payment activity
Now the agent has four different layers of financial reality instead of one giant context dump.
Lending or funding-operator stack
Plaid — financial-data infrastructure and integration operations
QuickBooks — business financials
Dun & Bradstreet — verified commercial identity and business-credit context
LoanPro — lending system of record
Moody’s — deeper credit and risk intelligence when economics justify it
Add orchestration around those systems and something much more interesting starts to emerge:
a funding agent that can gather context, identify missing information, inspect financial signals, enrich a business, summarize the file and prepare a human reviewer for the actual credit or lender-routing decision.That is materially different from:
Upload your bank statements to ChatGPT and ask if you're fundable.
One is infrastructure.
The other is chatbot roulette.
For a broader look at the surrounding AI stack, see 27 AI Tools Every Loan Broker Should Know in 2026 and the companion guide to AI Agents for Finance Operations.
For a borrower-facing next step, use the AI Funding Readiness Scorecard to frame what an AI-assisted funding workflow should collect before a human or lender reviews the file.
Finance MCP Security: Read-Only First, Autonomy Second
Finance is one of the worst places to adopt the philosophy:
Ship it and see what the agent does.
The MCP standard itself calls for strong controls around tool access and sensitive actions, while vendors such as Mercury intentionally restrict their servers to read-only access and Stripe explicitly warns about prompt-injection risk. MCP tools specification
A sensible deployment ladder is:
Read public documentation
Read sandbox data
Read production financial data
Generate recommendations
Prepare actions for approval
Execute low-risk reversible actions
Only then consider tightly controlled financial writes
Every level should have its own permissions.
Because “AI agent accidentally summarized the wrong invoice” is irritating.
“AI agent accidentally moved $80,000” is a meeting.
Are Finance MCP Servers Better Than APIs?
That is the wrong fight.
MCP does not replace the financial API.
It usually sits on top of or beside it.
Traditional APIs remain excellent when software knows exactly which operation it needs to perform.
MCP becomes valuable when an AI agent needs to:
discover available capabilities;
understand tool schemas;
select the appropriate operation;
combine data from multiple systems;
reason over the result;
continue a workflow conversationally.
LoanPro makes this distinction particularly well: APIs continue running the financial infrastructure while MCP becomes an AI-oriented interaction layer over that infrastructure. LoanPro: APIs and MCP
Think of the API as the plumbing.
MCP makes the plumbing understandable to the robot.
Preferably without letting the robot remodel the bathroom unsupervised.
What Finance MCP Servers Still Cannot Do
Even the best MCP servers do not magically solve:
bad source data;
undocumented lending policies;
regulatory requirements;
stale accounting;
weak permission design;
ambiguous business entities;
missing borrower documents;
bad prompts;
human judgment;
accountability.
Giving an AI agent five financial MCP servers does not turn it into Jamie Dimon.
It turns it into an AI agent with five financial MCP servers.
The difference remains economically important.
FAQs About MCP Servers for Finance
What are the best MCP servers for finance in 2026?
The strongest finance MCP servers include Ramp and Brex for finance operations, Mercury for business banking, Plaid for fintech infrastructure, QuickBooks for accounting, Stripe and PayPal for payments, Modern Treasury for treasury operations, LoanPro for lending, Dun & Bradstreet and Moody’s for credit intelligence, and Alpha Vantage or Morningstar for financial-market data and research.
What is an MCP server in finance?
A finance MCP server is a Model Context Protocol server that exposes financial data, tools or workflows to compatible AI systems. Depending on permissions, an AI agent may retrieve balances, analyze transactions, query accounting data, inspect loan information, access credit intelligence or execute approved financial operations.
Are there MCP servers for banking?
Yes. Mercury offers an official hosted MCP server with read-only access to business banking information. Ramp and Brex also expose significant financial-account and spend functionality through MCP.
Is there an MCP server for lending?
Yes. LoanPro has developed an MCP layer specifically for lending and servicing operations, with controls intended for regulated financial workflows.
Are there MCP servers for business credit?
Yes. Dun & Bradstreet exposes its Commercial Graph and business credit/risk intelligence through MCP. Moody’s also provides MCP access to proprietary credit ratings, entity intelligence, research and risk analytics.
Does Plaid have an MCP server?
Yes. Plaid offers a hosted Dashboard MCP server for production diagnostics and analytics plus local MCP tooling for development. Its official Dashboard MCP is primarily intended to manage and troubleshoot Plaid integrations rather than act as a general consumer bank-account chatbot.
Does QuickBooks have an MCP server?
Yes. Intuit maintains an official QuickBooks Online MCP server that connects compatible AI applications with QuickBooks entities and financial reports through OAuth.
Can ChatGPT use finance MCP servers?
Compatible remote MCP servers can be connected to AI platforms that support MCP or equivalent connector functionality. Availability, authentication and supported actions vary by server and AI platform.
Can an MCP server move money?
Some can expose write capabilities capable of creating or modifying financial objects; others deliberately operate read-only. Permission design matters enormously. Start with the least authority required for the workflow and keep human confirmation around consequential financial actions.
Are MCP servers safe for financial data?
They can be deployed safely, but MCP does not eliminate security risk. Use official servers when possible, scoped credentials, OAuth, least-privilege permissions, trusted clients, audit logging and human approval for sensitive operations. Even read-only access can expose highly sensitive financial information.
Are MCP servers replacing financial APIs?
No. MCP typically provides an AI-friendly interface to capabilities ultimately powered by APIs and other backend systems. APIs remain the operational foundation; MCP standardizes how AI agents discover and interact with them.
The Bottom Line
The real story about MCP servers for finance in 2026 is not that ChatGPT can suddenly talk to a bank.
It is that the financial stack is becoming agent-addressable.
Banking data can become context.
Accounting systems can become tools.
Commercial credit databases can become intelligence layers.
Loan-servicing platforms can become agent-accessible systems of record.
Payment infrastructure can become executable workflows.
That has enormous implications for lenders, brokers, CFOs, fintech founders and small-business operators.
The winning architecture is not:
AI replaces finance.
It is:
AI + verified financial systems + permissions + workflow controls + human judgment.
Build that correctly and the agent stops being a very articulate browser tab.
It becomes part of the financial operating system.
And that is where things start getting interesting.
This article is for informational and technology-comparison purposes only. MCP capabilities, access requirements and product features can change quickly. Financial, lending, credit and risk decisions should follow applicable policies, contracts, regulations and appropriate human review.
Comments