AI Affairs, home

Friday 2 October 2026

Corporate

Bloomberg launches Enterprise MCP for agents to access licensed market data

The service pairs financial data with field descriptions and reusable workflow procedures, while Bloomberg checks a client’s Data License rights before answering an agent request.

Modern architectural facade of the Bloomberg European Headquarters in London against a clear sky
Photo: DAVID HOLT, CC BY 2.0, via Wikimedia Commons (cropped)

Bloomberg launched Enterprise MCP on 29 September to give clients’ AI agents access to licensed data covering more than 100 million securities and over 50,000 fields, the company said in its announcement.

Key points

  • Bloomberg says Enterprise MCP provides an AI access layer for Data License Plus.
  • Bloomberg says metadata helps agents identify fields and interpret the data returned.
  • The initial release will include procedures for reviewing price exceptions and potential sanctions exposure, Bloomberg says.
  • Bloomberg says it checks a client’s licence rights before returning data to an agent.

Bloomberg attaches context to licensed fields

Enterprise MCP is an access layer for Bloomberg’s Data License Plus offering, according to the company. Bloomberg says the interface lets clients’ agents locate and retrieve licensed data through the Model Context Protocol, or MCP, rather than building separate connections for each category of content.

Bloomberg says its field metadata describes how a value is calculated and when it applies. Depending on the field, that context can include the exchange, currency, type of price, period and as-of date. Tony McManus, Bloomberg’s Global Head of Enterprise Data and Indices, said agents can reason through a question but still struggle with a position when the data supplied to them lacks that context.

Agents can search for a field by describing it in natural language, Bloomberg says. A separate entity-resolution function links names, tickers or descriptions to instruments and their issuing companies, according to the company. Bloomberg says one tool for finding securities and another for finding fields cover the content available through the interface.

Bloomberg says that content spans pricing and reference data, company information, economic data and alternative data across DL+ Live and the Qube Datastore. Bloomberg says new content becomes available to agents already connected to the interface without another integration.

Bloomberg Skills target trade reviews

Bloomberg is packaging reusable workflow procedures, which it calls Skills, alongside the data interface. The company says clients will be able to adopt and extend procedures for retrieving a universe of securities at a specified point in time, adjusting for corporate actions and examining revision histories.

The initial release will include Skills for finding unusual values among a group of tickers and reviewing trades outside accepted price-deviation thresholds, Bloomberg says. It will also include a procedure to assess securities and trades for potential sanctions exposure, according to the company. Bloomberg says it intends to add further workflow-focused Skills.

McManus said the difficulty for financial institutions has shifted from what an AI model can do to whether the data it receives is ready for use. Bloomberg says the combination of metadata, search functions and Skills is designed for research, portfolio, risk and operations work.

Data License rights govern agent requests

Clients can license the content domains they require under Bloomberg’s standard Data License rights, the company says. Bloomberg says it checks a firm’s entitlements before supplying data in response to any agent request. Requests are resolved against its Operational Datastore, while limits on individual tool calls restrict how many securities and fields can be returned.

Bloomberg hosts and manages Enterprise MCP, according to the company. Clients control the agents and applications connected to it, including their choice of models, prompts and workflow instructions. Bloomberg therefore supplies the data access layer and governs responses under its licences, while the client operates the applications that request the data.

Bloomberg plans to extend Enterprise MCP to real-time data for uses including intraday monitoring, checks before a trade and risk updates, the company says. It also says it is developing a separate way for Terminal subscribers to access Bloomberg data in desktop workflows, drawing on the governed access used by ASKB, the Terminal’s conversational AI interface.

Bloomberg participates in MCP governance

Bloomberg says its engineers are the principal authors of proposals to add enterprise governance controls to MCP. Those proposals would also allow servers to offer different sets of tools to different clients, according to the company.

Bloomberg says it established and leads the MCP Financial Services Interest Group. The company also says it holds seats on the board and Technical Committee of the Agentic AI Foundation, which oversees MCP as an open standard.

Topics: Agents, Enterprise adoption, Financial services