Snowflake Launches Cortex AI for Financial Services, Unifying Data and Agents with Managed MCP Server
Snowflake launches Cortex AI for Financial Services and a managed MCP Server to connect governed Snowflake data with external agents. Result: unified data and faster insights.

Snowflake launches Cortex AI for Financial Services
Snowflake announced Cortex AI for Financial Services - a suite built to help banks, asset managers, insurers, and fintechs unify data and deploy AI models, apps, and agents with strong security and compliance. Alongside it, Snowflake introduced a managed Model Context Protocol (MCP) Server in public preview to connect proprietary and third-party data in Snowflake to external agent platforms.
For finance teams, this means fewer integration projects, faster time to insights, and the ability to keep data governed while using best-in-class agents and models.
Why it matters for finance
- Unifies fragmented data so risk, compliance, and investment teams work from the same source of truth.
- Connects governed data to AI agents without custom plumbing - reducing time spent on one-off integrations.
- Supports regulatory needs with enterprise-grade security and governance controls built into Snowflake.
- Brings market data, research, and news into the same environment as your proprietary data for better context.
What's inside the Cortex AI suite
- Data Science Agent: An AI coding agent that automates data prep, feature engineering, model prototyping, and validation. Useful for quantitative research, fraud models, customer 360, underwriting, and compliance analytics.
- Cortex AISQL (public preview): SQL with AI functions for extraction and transcription across documents, audio, and images. Speeds up workflows like earnings transcript analysis, claims triage, and customer service summarization.
- Snowflake Intelligence (public preview): A conversational interface for business users to query both structured tables and unstructured documents in natural language. Reduces reliance on dashboards and ad-hoc analyst tickets.
- Industry data ecosystem: Structured data via Sharing of Semantic Views (generally available soon) from CB Insights, Cotality, Deutsche BΓΆrse, MSCI, and Nasdaq eVestment. Unstructured content via Cortex Knowledge Extensions (now generally available) from CB Insights, FactSet, Investopedia, The Associated Press, and The Washington Post.
MCP Server: connect AI agents to governed data
MCP offers a standard way for large language models and agents to interact with tools, APIs, and data. Snowflake's managed MCP Server lets external agents access Snowflake data and third-party shares while preserving security and governance.
- Interoperability out of the box: Connects Cortex Analyst and Cortex Search to remote agents through a standard MCP interface, unifying structured and unstructured retrieval.
- Broad agent support: Works with platforms such as Anthropic, CrewAI, Cursor, Devin, Salesforce's Agentforce, UiPath, Windsurf, Amazon Bedrock AgentCore, Azure AI Foundry, Glean, Kumo, Mistral AI, Workday, and WRITER.
- Fewer custom builds: Standardization cuts the need for bespoke integrations across teams and vendors.
New to MCP? The protocol details are available here: Model Context Protocol.
Practical use cases you can ship this quarter
- Market and portfolio research: Combine MSCI and Nasdaq eVestment with internal signals to generate screeners, summarize factors, and produce analyst-ready briefs.
- Quant workflows: Automate feature pipelines, backtests, and model iteration with Data Science Agent; keep reviews and approvals within your governed environment.
- Fraud and AML: Use AISQL to extract entities and patterns from unstructured reports and case notes; route suspicious activity summaries to case management.
- Claims and underwriting: Transcribe calls, parse documents, and auto-fill claim or underwriting fields; escalate only exceptions to adjusters and underwriters.
- Customer service and collections: Summarize interactions, propose next-best actions, and generate compliant responses straight from governed knowledge bases.
- Engineering productivity: With MCP, coding tools like Cursor can reference live data context to write more accurate production code tied to policy and schema.
What industry leaders are saying
- Anthropic: Securely connecting governed data to Claude removes a common blocker to production AI. Customers can apply advanced reasoning on both structured analytics and unstructured documents via Cortex tools.
- CrewAI: Multi-agent workflows need high-quality, secure data. The managed MCP Server provides the pipeline to analyze and act on governed data in Snowflake.
- Cursor: Coding assistants improve when they can access rich, live data context through a managed MCP environment.
- FactSet: AI-ready data products help clients unify and enrich data in modern cloud setups for risk, research, and long-term value creation.
- Ramp: Teams can query unstructured customer feedback in plain English and get instant answers, speeding up product decisions.
- Salesforce: Agentforce customers will be able to connect to Snowflake's MCP Server through AgentExchange for deeper cross-platform agent experiences.
Availability and notes
- Managed MCP Server is in public preview.
- Cortex AISQL's extraction and transcription are in public preview.
- Snowflake Intelligence is in public preview.
- Cortex Knowledge Extensions are generally available; Sharing of Semantic Views is expected to be generally available soon.
Some capabilities are preview-stage and may change before general availability.
How finance leaders can get started
- Pick 1-2 high-impact workflows: Examples: analyst research briefings, fraud case summarization, transcript analysis, claims intake.
- Consolidate data in Snowflake: Land core datasets and bring in third-party feeds through Marketplace shares, Knowledge Extensions, or Semantic Views.
- Stand up the managed MCP Server: Connect one agent platform your teams already use (e.g., Anthropic, Agentforce, Cursor) and test with governed sandboxes.
- Operationalize: Define approvals, logging, redaction, and data retention. Track precision, latency, and impact on cycle times and loss rates.
- Scale by template: Turn successful pilots into reusable playbooks for new desks, regions, or products.
Resources
Bottom line: Snowflake is bringing AI to where your data already lives and giving you a standard bridge to the agent platforms your teams prefer. If your priority is faster research, tighter risk controls, and leaner operations - without losing governance - this is worth a proof of concept now.