The Remesh MCP Server

Overview

The Remesh MCP server lets your team use Remesh from inside the AI assistant you already work in. Connect Claude or ChatGPT (or another MCP-compatible client) to your Remesh workspace, and you can browse studies, pull cited findings, compare segments, build discussion guides, and set up new conversations, all without leaving your AI tool.

Remesh is an official listing in both the Anthropic Connectors Directory (for Claude) and the OpenAI directory (for ChatGPT). In both, setup is a search, a click, and a sign-in with your Remesh account.

This feature applies to Live, Flex, and Video conversations. Learn about Remesh product offerings.

Included in This Article

• What Is MCP

• Which AI Tools Work With Remesh

• What You Can Do

• Remy or the MCP Server

• Citations

• Who Can Use It

• Security and Data

• Frequently Asked Questions

What Is MCP

MCP (Model Context Protocol) is an open standard that lets an AI assistant securely call an outside service through a defined set of tools. Remesh publishes its conversation, setup, and analysis capabilities as MCP tools. When you ask your AI assistant something about your research, it calls Remesh on your behalf and works with the live results.

In practice, this means you can type things like:

• "List the studies in my Remesh workspace that ended this quarter."
• "What did participants agree on most in the Q3 pulse conversation?"
• "Compare how managers and individual contributors talked about workload."
• "Pull five verbatims on onboarding friction, with citations."
• "Draft a follow-up discussion guide based on what we learned."

The AI does the connecting. You do the concluding.

Which AI Tools Work With Remesh

Step-by-step instructions for each client are in the Remesh MCP Setup Guide.

What You Can Do

Once connected, your AI client has access to roughly fifty Remesh tools in three groups.

Browse and set up. List workspaces, teams, folders, and conversations. Open a conversation and read its discussion guide, questions, and participant list. Create, duplicate, and schedule conversations. Build and edit sections and questions. Set the audience, including Remesh Recruit criteria and participant uploads. Export conversation data.

Analyze. Ask questions of one conversation or several at once. The AI client calls the same retrieval and statistics tools that power Remy's analysis in the platform: topics by agreement and by frequency, top responses, segment differences, sentiment distribution and examples, crosstabs, quantitative stats, correlation matrices, and relevant insights for a question. Every analysis answer carries citations back to Remesh.

Run Remy flows. Kick off Remy's discussion guide creation (AutoDG) from a research objective and continue the conversation turn by turn. Ask Remy platform help questions. Retrieve the results of a prior Remy run.

What it deliberately cannot do. The MCP server has no tool to delete a conversation, moderate a live session, or manage users, permissions, billing, or workspace settings. Those stay in the Remesh platform.

Remy or the MCP Server

Same data, two front doors. Both run on the same studies, the same permissions, and the same citations. The difference is where you work.

Citations

Analysis answers include the same numbered citations you see in the Remesh dashboard. Each [N] marker links to the source view in Remesh, so you can click through and verify any claim against the underlying responses. Where a citation refers to a single participant response, it also carries that response's identifier. Where it refers to an aggregate (a topic, a segment, a summary), it links to the relevant analysis view.

Client behavior varies. Claude and ChatGPT render citations as clickable links. Some clients show the bare [N] markers with the citation list underneath.

Who Can Use It

Three things have to be true before a person can use the Remesh MCP server. All three are controlled by your Workspace Administrator in Manage Workspace > Settings.

1. Generative AI is enabled and approved for the workspace (Generative AI tab). The MCP server is built on Remesh's AI features and is unavailable without them.
2. MCP access is enabled for the workspace (MCP Access tab). The administrator acknowledges the MCP terms and turns on "Enable MCP server access for this Workspace."
3. MCP access is enabled for the person's team(s) (MCP Access tab, team checklist). This step is easy to miss. The workspace toggle alone does not make any data reachable. Each team that should be able to use the MCP server must be checked, and that team must also have Generative AI enabled.

If you connect successfully but every request comes back with "no team memberships" or "not authorized," step 3 has not been completed for your team. Ask your Workspace Administrator to enable MCP access for your team.

Full instructions are in the Setup Guide.

Security and Data

The short version, for researchers. The detailed version for your security team is in the Remesh MCP Security and Data Handout, available from your account team.

• You sign in as yourself. Every connection uses OAuth 2.0 with your Remesh login. There are no shared API keys.
• Permissions are inherited, never expanded. Every request is checked against your Remesh workspace and team membership before it runs. Disabled teams' data is unreachable.
• Data flows to the AI client you choose. What the AI reads travels encrypted to Claude, ChatGPT, or whichever client you connected. That leg is governed by your organization's agreement with that AI vendor. Remesh adds no new subprocessors.
• The MCP layer keeps no copy of your research data. It holds short-lived session state (which workspace you selected, for up to 24 hours), rate-limit counters, and operational logs, all inside Remesh's existing production infrastructure.
• Rate limited. Requests are limited per workspace and per user to protect platform performance.
• Auditable and reversible. Remesh records who enabled, requested, and revoked MCP access, and when. A Workspace Administrator can turn access off for a team or the whole workspace at any time; existing connections stop working immediately.

Frequently Asked Questions

Which AI tools are supported? Claude (official Anthropic Connectors Directory listing), ChatGPT (official OpenAI directory listing), Microsoft Copilot Studio (added by URL), and any custom agent that supports MCP over HTTPS with OAuth 2.0. See the Setup Guide.

Is Remesh listed in the Microsoft directory? Not at this time. Copilot Studio makers can still add the Remesh MCP server to an agent by URL.

Do I need to be an administrator? To connect and use the MCP server, no. You need a Remesh account with team membership on a team that has MCP access enabled. Enabling MCP access for the workspace and for teams requires a Workspace Administrator. On Claude Team/Enterprise and ChatGPT Business/Enterprise plans, your AI vendor's own administrator may also need to make the Remesh app available to your organization.

Does connecting a client give it more access than I have? No. The server enforces your existing Remesh permissions on every request and cannot exceed them.

Can the AI change or delete my studies? It can create and edit conversations, discussion guides, sections, questions, and audience settings, within your permissions. It cannot delete a conversation, moderate a live session, or change users, permissions, or workspace settings.

Is our data used to train AI models? On the Remesh side, no. What happens once data reaches your AI client is governed by your organization's agreement with that vendor (for example, Anthropic's or OpenAI's enterprise terms). You choose the client, so your existing enterprise AI terms apply.

Do analysis answers include sources? Yes. Answers carry the same [N] citations you see in the Remesh dashboard, linking back to the source view.

Can I ask about several studies at once? Yes. Analysis tools accept multiple conversation IDs (up to 10 per call), so you can ask cross-study questions in a single prompt.

I connected successfully but nothing works. Every request says no team memberships or not authorized. MCP access is enabled for your workspace but not for your team. Ask your Workspace Administrator to open Manage Workspace > Settings > MCP Access and check your team. The team must also have Generative AI enabled.

I am getting rate-limited or slow responses. Requests are limited per workspace and per user, and Remy flows have a small concurrency cap per workspace. Space out large batches and retry.

Does this replace Remy? No. Remy is the in-platform experience with zero setup. The MCP server brings the same grounded data to the AI tools your team already uses. Same data, two front doors.

I found an issue or have feedback. File a support ticket or contact your account team.