Bloomberg launches Enterprise MCP AI access layer for Data License Plus
Bloomberg announced in October 2025 an Enterprise MCP server that lets client AI agents retrieve Data License Plus content, including reference, pricing and corporate actions data across roughly 100 million securities, inside the client's own applications under existing Data License entitlements. Bloomberg also ships Skills that let agents discover and request licensed fields. The announcement discloses no separate MCP pricing or seat change.
Where the rake sits
Bloomberg keeps the rake and it is owned: the substrate is Bloomberg's, the semantic layer that makes it agent-addressable (the MCP Skills) is Bloomberg-built, and the client pays a Data License Plus fee that is not shared with a partner. MCP as a protocol is open, but the endpoint, entitlements and content stay inside Bloomberg's perimeter, so no external party is doing enrichment work that could walk.
What happened
- Bloomberg launched an Enterprise MCP (Model Context Protocol) server exposing Data License Plus content to client AI agents
- The endpoint reaches Bloomberg's reference, pricing and corporate-actions data across ~100 million securities
- Access is gated by existing Data License entitlements; the announcement discloses no separate MCP pricing, seat count or bundling change
- Bloomberg ships 'Skills' that let client agents discover and retrieve licensed fields via MCP
- Data License Plus is the same feed franchise that sits behind the Bloomberg Terminal (~US$15bn 2024 revenue, ~26,000 employees)
Who is involved
Bloomberg is an American financial, software, data, and media company headquartered in Midtown Manhattan, New York City.
The reading
Bloomberg does the enrichment: it curates the underlying securities data, adds AI-ready metadata and semantic context, and ships Skills that let client agents discover and retrieve licensed content. The MCP layer is Bloomberg-built on top of a Bloomberg-owned substrate.
The buyer's decision is which security a client's AI agent should reason over inside the client's own workflow — pricing, reference data, corporate actions across ~100 million instruments — and the substrate is worth paying for because that same coverage already sits behind the Terminal franchise and cannot be reconstructed from public sources.
What crosses is retrieval of licensed content into the client's enterprise AI application under existing Data License terms; the corpus itself, the identifiers and the semantic layer stay inside Bloomberg's perimeter. Delivery through an MCP endpoint does not change what is being sold — it is a sale of the data, not of a wrapping product.
Not determinable; the press release states no pricing, no seat count and no bundling terms relative to a standard Data License Plus contract.
Why it matters
Shipping the MCP Skills under existing Data License entitlements is the move that keeps the rake owned: Bloomberg built the semantic layer that makes the substrate agent-addressable itself, so no integrator sits between the ~100 million-security corpus and the client's agent. The risk is pricing, not architecture. A per-seat Data License fee against agent traffic that could pull the same field thousands of times a day is where under-capture would show up, and the announcement discloses no new meter.
The argument this deal tests: You have an AI agent. Great. Does it have the data?
Related deals
- PitchBook becomes data partner for OpenAI's ChatGPT for Financial Services, delivering firmographic data via Essential connector2026-09-10
- Swiggy to launch Builders Club, opening 3 MCP servers and 18+ API tools to external developers on AWS2026-04-27
- LSEG and Snowflake announce expanded five-year collaboration to connect financial data with cloud-native AI workflows2026-09-30
- Quartr becomes a data partner to OpenAI's ChatGPT for Financial Services2026-09-10
Sources
- bloomberg.com — primary
- en.wikipedia.org — party background
Announced 2026-09-29