AnswerThis MCP Server: Research Setup and Permissions

Website image

Ryan McCarroll

10 min read

AnswerThis MCP connector
AnswerThis MCP connector

An MCP server lets an AI assistant request information or actions from an external service. AnswerThis uses this connection for paper search, saved research libraries, and Writer drafts. Public paper search does not require a connected AnswerThis account. Library functions and draft creation or editing require your account and the relevant permissions. The connection brings research tools into a compatible assistant, but you still need to check the sources behind its answers.

TL;DR

  • An MCP server exposes external tools and information to a compatible AI assistant.

  • AnswerThis public paper search works without connecting an AnswerThis account.

  • Library access and Writer draft actions require account permissions. Full-text availability depends on the source.

  • Use AnswerThis’s setup page for connection details. A public product-page address is not a server address.

  • Check retrieved references and approve changes to saved research before authorizing them.

What the AnswerThis MCP server does

The AnswerThis MCP overview documents paper search by topic, DOI, title, or author. With an account connected and the required permissions approved, you can also browse library folders, save selected papers, and read accessible library text.


AnswerThis MCP research workflow


The writing functions let an assistant create a Writer draft with linked citations or edit a saved draft. Saving a draft is an explicit action. Finding papers and changing a persistent document are different operations, even when both happen in the same conversation.

AnswerThis MCP suits researchers who want to discover papers and continue working with saved sources and drafts inside a compatible assistant. As with any research source, open the full text and read the cited passage before you rely on a claim.

The full AnswerThis research workflow also offers document uploads, import-status checks and a dedicated evidence-passage tool.

What is MCP, or Model Context Protocol?

MCP is an open standard for connecting AI applications to external information and tools. It supplies a shared interface rather than a research database or a new language model.

Anthropic introduced MCP on 25 November 2024. The official MCP architecture distinguishes the host application, a client component that communicates with a server, and the server providing capabilities. The assistant’s application decides when and how to use those capabilities.

What is an MCP server?

An MCP server is software that exposes selected information or actions to a compatible application. A search function can return publication records. A writing function can create or update saved material if the connected account allows it.


MCP Server Overview

An illustrative research workflow is:

  1. You ask the assistant to find papers on a topic.

  2. The application calls a search function exposed by the server.

  3. The service returns records or available text.

  4. The assistant uses the results to prepare its response.

The protocol does not guarantee that this sequence occurs for every answer. Inspect tool activity and source links where the application makes them available.

Before you start

Prepare the access needed for the task, rather than assuming every operation needs the same account connection.

  • For discovery: Choose an assistant that supports the connection method described in AnswerThis’s current setup instructions.

  • For saved research: Use the AnswerThis account containing the library or drafts you intend to work with, and review the permissions requested.

  • For connection details: Open the AnswerThis MCP setup page for current client instructions, account requirements, and the remote-server address.

Do not paste the public product-page address into a server-address field. Copy the connection address from the setup instructions. This guide does not publish an unverified address or assume a particular subscription entitlement.

The setup page is the starting point for current product-specific details. If it requires sign-in, use your account to view the instructions. Public paper search works without connecting an AnswerThis account; your client’s own sign-in and data settings still apply.


AnswerThis MCP setup

Supported assistants: distinguish client support from product support

An assistant supporting MCP does not establish that a particular connection has been tested, appears in its directory, or works with every account configuration.

Claude

  • Official connection guidance: Supports custom remote MCP connectors, with different setup responsibilities for individual and organization accounts

  • What to check for AnswerThis: Use the server address and authentication instructions supplied by AnswerThis

ChatGPT

  • Official connection guidance: Documents adding custom MCP servers, subject to account and workspace policies

  • What to check for AnswerThis: Confirm that the current AnswerThis instructions support your intended client and workflow

Claude’s current connector guide explains individual and organization setup. OpenAI’s connection documentation describes its own MCP setup and testing workflow.

Avoid choosing a subscription solely from a generic MCP support statement. Check the requirements shown by both services for the connection you plan to use.

Set up a research connection in Claude

The following steps combine Claude’s documented connector controls with AnswerThis’s published capability boundaries. Follow them in order and confirm each step in your client.

1. Obtain the connection details

Open the AnswerThis setup page and read the instructions for your chosen assistant. Copy the remote-server address and note the authentication required for your intended operations.

Expected result: You have a service-provided connection address, not a guessed address based on a website URL. If the page does not provide instructions for your client, resolve that before proceeding.

2. Add the connector in Claude

For the individual setup documented for Pro and Max, open Customize > Connectors, select + Add, then Add custom connector. Enter a name and the supplied remote-server address, then select Continue.

Review the detected authentication settings and continue through the configuration. Use the authentication method specified by AnswerThis. Do not supply invented credentials or choose no sign-in merely to bypass an authorization problem.

For organization accounts, an authorized administrator may need to add the connector through Organization settings > Connectors before members connect individually.

Expected result: The connector appears in Claude’s configured list. An entry in that list is not yet proof that a search works.

3. Review account permissions

Connect your AnswerThis account when the requested operation requires it. Check that you selected the account containing the intended research materials.

Review library and writing access separately. Reject unexpected permissions rather than accepting them just to finish setup. Keep write actions subject to review instead of granting blanket approval.

Expected result: The authorized functions match your intended task. Do not assume that authorization for reading also includes editing.

4. Enable and test search

In a conversation, open the + menu and Connectors, then enable the configured connection. Start with a known public paper identified by DOI or title.

Compare the returned title, authors, and identifier with the original publication record. Inspect the tool result where available. A fluent response alone does not establish that the connector ran.

Expected result: You can identify the retrieved paper and the source of the returned information.

5. Test a reviewed save or draft action

After checking the search result, ask the assistant to propose saving a selected paper or creating a draft. Review the selection or proposed text before authorizing the action.

Open the relevant library or Writer draft afterward and inspect the saved result. Do not treat an assistant’s statement that it saved something as sufficient confirmation.

Expected result: The intended item exists in the correct account, without unrelated changes.

Research functions and their boundaries

Search by topic, DOI, title, or author

  • Account boundary: Public search does not require connecting an AnswerThis account

  • What you should verify: Publication identity and relevance

Browse folders and save papers

  • Account boundary: Connected account and required permissions

  • What you should verify: Correct folder, paper, and account

Read accessible library text

  • Account boundary: Connected account and required permissions

  • What you should verify: Whether the returned material includes the text needed

Create or edit Writer drafts

  • Account boundary: Connected account and writing permissions

  • What you should verify: Proposed changes and the saved document

These descriptions are capabilities, not technical tool identifiers. The functions exposed in your client and the permissions approved for your account determine what is available.

Full-text access remains source-dependent. If only metadata or an abstract is available, ask the assistant to preserve that limitation in its response rather than describe it as a full-text review.

Example requests

These are suggested instructions, not transcripts of a completed demonstration.

Find an identifiable paper: Ask the assistant to search for a DOI or title, return the matching publication details, and state what material it retrieved.

Explore a topic: Ask for relevant papers, the search terms used, and a distinction between retrieved information and the assistant’s interpretation.

Save a reviewed selection: Ask to see the proposed papers and destination folder before approving a save.

Revise a selected draft: Ask for proposed changes using selected sources, with no saved edits until you approve them.

For a formal review, retain more than the conversational request. Record the actual database and platform, search date, executed queries and filters, and result counts where available. An assistant’s prompt is not a substitute for a reproducible search strategy.

Troubleshooting the connection

The connector does not appear. Confirm it was added to the intended account or organization. On a managed account, ask the administrator whether the connection has been enabled for members.

The server address is rejected. Compare it with the current AnswerThis setup instructions. Do not substitute the public product page or alter the address by guessing.

Search works, but library or draft actions do not. Check account connection and the permissions for the requested operation. Public search and account-linked functions have different requirements.

A paper is found, but text is unavailable. Discovery does not guarantee full-text access. Inspect the source and available material instead of asking the assistant to reconstruct missing text.

The assistant answers without visible retrieval. Confirm the connector is enabled for the conversation and ask it to use the relevant search function. If activity cannot be verified, do not describe the response as a checked database search.

Data, privacy, and source checking

The AnswerThis data handling guide says uploaded analysis documents are stored temporarily, removed periodically, not shared with other users, and not used to train its AI models. It does not specify an exact retention period or comprehensively describe MCP query and tool-result handling.

Do not extend those document statements into a blanket promise about every connected assistant. Review both services’ terms and your institution’s requirements before sharing confidential research or participant information. Library access does not itself establish permission to transmit licensed text to an external service.

A retrieved reference also needs checking. Open the publication record, inspect the relevant passage, and confirm that it supports the claim being drafted. Check corrections or publication notices at the source.

AnswerThis says connected apps can be revoked from its service. Revoking access is different from deleting material already stored elsewhere. Check deletion arrangements separately when that matters to your research.

FAQs

What is an MCP server?

An MCP server exposes an external service’s information or actions to a compatible AI application. It supplies capabilities such as search or document editing, not a guarantee of accurate answers.

Do I need to connect an AnswerThis account to search papers?

AnswerThis says public paper search is available without connecting an AnswerThis account. Library and draft operations require the account and relevant permissions, while your assistant may have its own connection requirements.

How do I connect Claude to AnswerThis?

Get the remote-server address and authentication instructions from AnswerThis’s setup page, then follow Claude’s custom-connector instructions. Enable the connector and verify a known paper before using saved-library or writing actions.

Can I use AnswerThis MCP in ChatGPT?

ChatGPT supports custom MCP connections, so you can connect AnswerThis there too. Follow the current AnswerThis client instructions and applicable ChatGPT account and workspace policies.

Does finding a paper mean full text is available?

No. Full-text availability depends on access to the source. Identify whether the tool returned publication metadata, an abstract, or accessible text before describing what the assistant reviewed.

Can the assistant edit my AnswerThis drafts?

AnswerThis documents creating and editing Writer drafts with a connected account and writing permissions. Review proposed edits and inspect the saved draft after authorizing a change.

Is information shared through MCP used for AI training?

According to AnswerThis’s guide, uploaded user documents are not used to train its AI models. That statement does not determine every MCP query’s handling or another assistant provider’s policies. Review both services before transmitting sensitive material.

Start with a reversible research task

Begin with a public record you can verify, then add account-linked operations only when needed. A successful connection is not just an enabled connector: it is a retrieved source you can inspect and, for write actions, a saved result you have checked.