Bloomberg launches Enterprise MCP for AI data access

Bloomberg has launched Enterprise Model Context Protocol, an AI access layer for its Data Licence Plus service aimed at financial institutions using AI agents with Bloomberg data.

The system gives client applications a standardised way to discover, interpret and retrieve licensed Bloomberg data across more than 100 million securities and 50,000 fields.

Enterprise Model Context Protocol, or MCP, is designed to address a problem that has emerged as firms test generative AI in financial services: data may be accessible, but AI systems often lack the context to understand what a figure represents, how it was calculated and whether it is suitable for a specific task.

Bloomberg is packaging the service with metadata that explains fields and qualifiers such as exchange, currency, price type, period and as-of date. It is also exposing that metadata through semantic search tools, allowing agents to search for fields and instruments in natural language rather than through fixed mnemonics or manually maintained identifier lists.

The service sits on top of Bloomberg's Data Licence Plus offering and brings together pricing and reference data across asset classes, company fundamentals and structure, economics and alternative data, including material from DL+ Live and the Qube Datastore.

Data context

For financial institutions, the issue is not simply pulling a number from a database. Without context, an AI system may be unable to tell which currency applies, what type of price has been provided or when the value was recorded.

The aim is to let users ask for information in plain language and receive data with enough context to support research, portfolio work, risk checks and operations tasks. Bloomberg also said firms should be able to move from AI pilots to production workflows more quickly if those controls and definitions are embedded in the data retrieval process.

"As generative AI adoption in financial services has grown, the bottleneck facing financial institutions has shifted from model capability to data readiness," said Tony McManus, Global Head of Enterprise Data and Indices at Bloomberg.

"We've seen agents that reason well but still can't answer a basic question about a position, because the data reaching them carries no indication of what it actually represents. On its own, a last price doesn't say what currency it's in or what kind of price it is. Connections were never the hard part. Readiness is, and that's what we built Enterprise MCP to solve," said McManus.

Workflow tools

The product includes task-based tools rather than separate interfaces for each asset class. A single security discovery tool and a single field discovery tool cover the available content set.

The first release also includes reusable procedures, described by Bloomberg as Skills, for point-in-time universe retrieval, corporate action adjustment and revision history. Initial workflow functions include identifying outliers in a group of tickers, reviewing trades that breach accepted price deviation thresholds and assessing securities and trades for potential sanctions exposure.

That focus suggests Bloomberg is positioning the product closer to day-to-day operational work rather than limiting it to experimentation by data science teams. Clients will also be able to adopt and extend those procedures within their own workflows.

Controls and access

Bloomberg will host and manage Enterprise MCP, while clients will control the AI agents and applications that connect to it, including models, prompts and instructions. The company will validate a firm's entitlements before returning data for any agent request under standard Data Licence rights.

Requests are resolved against Bloomberg's Operational Datastore, and tool-level limits cap the number of securities and fields returned in a single call. Those controls are likely to matter for regulated firms seeking auditability around how AI tools access market and reference data.

Bloomberg also said its engineers are involved in developing MCP as an open standard. The company is the primary author of proposals intended to add enterprise governance controls to the protocol and to allow servers to present different toolsets to different clients. It also leads the MCP Financial Services Interest Group and holds seats on the Board and Technical Committee of the Agentic AI Foundation, which stewards the standard.

The launch reflects a broader shift in finance as data vendors and software groups seek to build governed routes between proprietary datasets and generative AI systems used inside banks, asset managers and other market participants. Enterprise MCP is available as an enterprise offering, with clients able to license specific content domains under standard Data Licence rights.