MCP, in the assistant you already use

Ask in the assistant you already use.

A short list from the national docket comes back. Connect Claude, ChatGPT, Cursor, or the agent you prefer. What filed, which cases match the mandate, what is on the claims register: without leaving the conversation, and without opening PACER case by case.

How to connect What this does for you

What this does for you

The desk stays in one conversation. The market answers.

MCP is not a second product. It is how the custom report, the claims register, and the court PDF arrive in the assistant the team already uses. Names, inputs, and JSON-RPC live in the developer reference.

See what filed that matches

A 363 in Delaware and a disclosure statement in Texas land on the same list. PACER only shows those events if you already knew the case. Ask for the custom report we built. The matching cases come back.

Open the claims register on a named case

Creditor, filed amounts, status, proofs of claim. The register is in the same conversation that found the filing.

Get the papers

Pull the court PDF from the same conversation that found the filing. The motion is in hand while the auction is still open.

Run your custom report from the chat

The report we built around the mandate is a question you can ask. Same list as the morning email, on demand, in Claude, ChatGPT, Cursor, or the agent you prefer. Custom reports are the product. MCP is one way they arrive.

A morning in the assistant

A question you would actually ask. A short list from the courts we cover.

“Run my healthcare Chapter 11 screen for the last 30 days.”

The custom report from the trial: court, chapter, filing date. The screen that used to take a morning in PACER.

“For the largest one, show the claims register and pull the three biggest proofs of claim.”

The creditor list, filed amounts, and the claim PDFs in hand. Before the register goes stale.

“Run my sale-motion report for the last two weeks, then pull the motions.”

The 363 motions nationally in the window, documents in hand. Before the auction, not after the sale order.

“Who bought claims last week, and in which cases?”

Market intelligence on claims trades: we publish a weekly report on claims trades from across the US. That data is available back to 2014. Ask for last week, or a buyer, or a case.

Illustrations written by us, not live output. Results depend on the filings in the window and the custom report on your subscription.

What you type / what comes back

One question. A short list from the courts we cover.

You type

Sale motions in healthcare, last 14 days, liabilities over $25 million.

What comes back

Matching cases from the custom report on your account

  • D. Del. · Ch. 11 Motion to sell property free and clear of liens
  • S.D.N.Y. · Ch. 11 Motion to sell substantially all assets
  • N.D. Ill. · Ch. 11 Motion to approve bid procedures and a sale

Illustration, not live output. Debtor names omitted. The rows on your account are the cases that matched your mandate that day.

Connect

One URL. Authorize once. Then ask for the report we built.

MCP is included with every TrollerBk subscription. We put your custom reports on the account, then they show up as tools in Claude, ChatGPT, Cursor, or the agent you prefer. After you log in, your API token is under Your Account → API Token. OAuth authorizes the assistant without pasting that long-lived token into the chat. Endpoint: https://mcp.trollerbk.com/mcp.

  1. In Claude, open Settings → Connectors and choose Add custom connector.
  2. Name it TrollerBk and paste the URL https://mcp.trollerbk.com/mcp.
  3. Click Connect. A TrollerBk authorization page opens; paste your API token from Your Account → API Token and click Authorize.
  4. Start a new chat and ask for the report from your trial. If we built you a sale-motion screen: “Run my sale-motion report for this week.”

Claude receives a short-lived OAuth token, never your subscription token.

  1. In ChatGPT, open Settings → Connectors (developer mode may need to be enabled) and create a new connector.
  2. Server URL: https://mcp.trollerbk.com/mcp. Authentication: OAuth. ChatGPT discovers the rest from https://mcp.trollerbk.com/.well-known/oauth-authorization-server.
  3. When prompted, paste your TrollerBk API token on the authorization page.
  4. Enable the connector in a new conversation and ask for the report from your trial.

If the connector appears but ChatGPT will not use it, start a fresh conversation with the connector toggled on. The developer reference has the troubleshooting list.

Add TrollerBk to your mcp.json (Cursor) or claude_desktop_config.json:

{
  "mcpServers": {
    "trollerbk": {
      "url": "https://mcp.trollerbk.com/mcp"
    }
  }
}

Leave out the headers and the client runs the OAuth flow in your browser. Prefer a static credential for a headless agent? Add "headers": { "Authorization": "Bearer <your-api-token>" } instead.

Any client that speaks MCP over HTTP can connect to https://mcp.trollerbk.com/mcp. OAuth 2.1 with PKCE, or a Bearer token from Your Account → API Token.

JSON-RPC methods, tool schemas, curl examples, and OAuth discovery are in the developer reference.

In the assistant the desk already uses

Connect Claude, ChatGPT, Cursor, or the agent you prefer. The portal remains for docket refresh, documents, and the claims register. MCP is the same corpus inside the conversation.

No chatbot on this website

TrollerBk does not run a model. Questions live in your assistant's session. TrollerBk returns structured filings, claims, and documents.

Subscriber-only, same corpus

There is no free tier and no anonymous prompt box. A custom report against the national docket is the product. MCP is one way it arrives. Free single-case lookups live on the BankruptcyObserver MCP.

MCP questions

The Model Context Protocol is the open standard Claude, ChatGPT, Cursor, and other agents use to call outside data. TrollerBk publishes the national bankruptcy corpus that way. You stay in the conversation you already have. The docket answers. No export ritual, no second website to live in.

No. TrollerBk does not run a model or a chatbot on trollerbk.com. You connect the assistant you already use and trust. It calls TrollerBk with your subscriber credential. Questions and results stay inside that session.

Yes. Every TrollerBk plan includes MCP. There is no connector fee. Signup starts a trial: we build the custom reports around your mandate, then those reports appear as tools in the assistant. Your API token is under Your Account → API Token after you log in. OAuth lets an assistant authorize without ever holding that long-lived token.

Claude (desktop and web connectors), ChatGPT connectors, Cursor, mcp-remote, and any client that speaks MCP over HTTP with OAuth 2.1 or a Bearer token. Protocol details live in the developer reference. Want a different assistant? Tell us and we will set it up.

It gets the same corpus as your subscription: market screens, dockets, claims registers, documents, and your custom reports. It does not get billing or account administration. OAuth issues short-lived access tokens with refresh rotation. You can revoke them at any time.

BankruptcyObserver is a lookup: start with a case you already know and go deep, with a free tier for exact-name searches. TrollerBk starts from the court event feed, links each line to a case, a docket, and a type, then you go deep. It is subscriber-only, built for investors and advisors.

The first question is the report we built for the trial.

Tell us the mandate. We build the custom reports, they appear as tools in the assistant your analysts already use, and a short list comes back from the courts we cover.

See pricing Custom reports