Bloomberg Law

Search federal and state court dockets, retrieve filings, and create alerts and tracks.

Opens Agent Studio, where connecting is one click. The connector URL below works in any MCP client.

Paste it into any MCP client. Setup docs

What the Bloomberg Law connector does

The Bloomberg Law MCP server connects Bloomberg Law's federal and state court docket collection to Claude, ChatGPT and any MCP-compatible agent. Its 17 tools search dockets, retrieve metadata, entries and filing PDFs, predict federal civil outcomes, and create alerts and tracking subscriptions. Sign-in is OAuth against a Bloomberg Law subscription with connector access enabled.

Verified connector

Listed by Anthropic as a partner connector in its Connectors Directory.

Connection checked by Agentman on .

Anthropic states this reflects the level of review a connector received, not a security audit.

Bloomberg Law tools (17)

  • BLAW__docket_download_entry
  • BLAW__docket_execute_alert_creation
  • BLAW__docket_execute_filing_request
  • BLAW__docket_execute_track_creation
  • BLAW__docket_execute_update_dockets
  • BLAW__docket_get_docket_cost
  • BLAW__docket_get_entries
  • BLAW__docket_get_filing
  • BLAW__docket_get_metadata
  • BLAW__docket_get_outcome_prediction
  • BLAW__docket_get_pleading_cost
  • BLAW__docket_preview_alert_creation
  • BLAW__docket_preview_filing_request
  • BLAW__docket_preview_track_creation
  • BLAW__docket_preview_update_dockets
  • BLAW__docket_search
  • BLAW__docket_visualize_search_results

Tool names from Anthropic's directory listing. This server requires sign-in, so we could not read tool descriptions or parameter schemas.

Limits

  • Dockets only, and Bloomberg says so explicitly. Case law research, non-docket legal research, judge analytics, litigation analytics and broader company intelligence are all directed back to Bloomberg Law's own tools. This is a negative enumeration — a published list of what the connector cannot do — which constrains the surface more usefully than a feature list does.
  • Filing requests and docket updates cost money, with no published rate card and no documented idempotency key.
  • Entitlements gate everything. You reach only the content your own subscription covers, so two users on the same query can get different results.
  • The licence prohibits building a retrieval corpus from returned content, and prohibits prompt use generally outside the consent this connector represents.
  • An automated bot is not a licensed User under the subscription terms. Unattended pipelines on a shared service account are the arrangement most clearly outside the definition.
  • Volume limits are discretionary, not numeric. Bloomberg judges systematic use by its own good-faith assessment of quantity and frequency, with no threshold published to stay under.
  • We could not read tool schemas or safety annotations. The endpoint returned 401 to an anonymous request, so no parameter-level detail appears here and the read/write labels are Bloomberg's rather than the server's machine-readable declaration.
  • Bloomberg disclaims what Claude generates. Its documentation states that a summary Claude produces from an uploaded filing "is generated by Claude, not Bloomberg Law", and that visualizations built from exported data "are not produced or validated by Bloomberg Law". It advises validating important information and exercising professional judgment. Outcome predictions are Bloomberg's, but the narrative around them may not be.
  • analytics:write and mcp:tools are unexplained. Both appear on Bloomberg's authorization server and neither maps to any listed tool. The connector's own resource descriptor advertises only mcp:read and mcp:write, and we treat the wider list as the vendor's general API vocabulary rather than as evidence about this connector.
  • The tool count is a floor. Bloomberg states additional capabilities may be added and that new tools will then appear automatically.

Frequently asked questions

Does the Bloomberg Law connector return the full text of filings?

Yes, for filings you are entitled to. Two tools reach document content: docket_download_entry returns a docket entry as a formatted PDF, and docket_get_filing fetches a docket entry's PDF content for summarization or analysis. That is full document text crossing into the model's context, not a citation or an abstract, which makes the licensing terms below the operative constraint.

Does Bloomberg Law's licence allow feeding its content to an AI model?

Not in general, and the connector is the licensed exception. Bloomberg Industry Group's subscription terms bar using Product content in prompts or for training, tuning or incorporating in any Language Model without its written consent, which it may withhold at sole discretion. Bloomberg operating this connector itself is what makes connector use sanctioned.

What OAuth scopes does the Bloomberg Law MCP server request?

Two, and they draw a real boundary. The server's RFC 9728 descriptor advertised mcp:read and mcp:write on 2026-08-23, and the 401 challenge repeated both. Because every state-changing tool is an execute variant, the write scope maps onto exactly the operations that spend money or create subscriptions, which a read-only grant would exclude.

Can the Bloomberg Law connector spend money without asking?

It should not, because Bloomberg gates spending behind a preview step. Four write tools each require a matching preview call first, and Bloomberg documents that Claude previews eligibility and any applicable cost and requires user confirmation before executing. Two dedicated tools, docket_get_docket_cost and docket_get_pleading_cost, exist purely to report price before you commit.

Can the Bloomberg Law connector delete anything?

No. No tool among the 17 carries a delete, remove, cancel or unsubscribe verb, in either Anthropic's listing or Bloomberg's own published table. The write tools only create alerts and tracks, request filings, and refresh dockets. Removing a track or alert you created is not available through this connector and must be done in Bloomberg Law directly.

What personal data can the Bloomberg Law connector read?

Names of real litigants, unavoidably. Court dockets identify parties, and docket_get_metadata returns court information, case details, parties and attorneys by design. Filing PDFs carry whatever the filers put in them. That is the corpus working as intended, but it means real personal data about identifiable people enters your agent's context on ordinary queries.

Does the Bloomberg Law connector cover case law research?

No. Bloomberg's documentation scopes the connector to dockets and directs you to use Bloomberg Law research tools directly for case law research or non-docket legal research. Judge analytics, litigation analytics and broader company intelligence are also excluded. Every one of the 17 tool names begins with a docket prefix, which matches that boundary exactly.

What subscription do I need for the Bloomberg Law MCP connector?

An active Bloomberg Law subscription with MCP connector access enabled. Bloomberg states that if you are unsure whether your account has access you should check with your Bloomberg Law administrator or account representative. Entitlements then apply per user, so you can only reach the Bloomberg Law content your own subscription already covers.

Sources

Use it in an agent

Put Bloomberg Law to work.

Connect Bloomberg Law once and your agents call these tools on their own: on a schedule, in a workflow, with nobody at the keyboard.

Call (650) 285-1019Our AI receptionist answers.