Remesh MCP server icon

Remesh

by Remesh

HIPAA CompliantSOC2 ReadyISO 27001 Ready
Analytics52 tools

Design, build and analyse Remesh research conversations from an AI agent. Anthropic lists 52 tools — 29 read, 21 write, 2 delete. The read surface returns individual participant verbatims, and nothing can send a question into a live session.

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.

Connect Remesh via MCP

https://live.remesh.chat/mcp

Works in any MCP-compatible client. In Agentman Studio it is one click — no config file to edit.

Remesh Tools & Capabilities (52)

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

Limits

  • Remesh publishes no tool names. Its MCP article describes three capability categories rather than an enumeration, so all 52 names matched semantically only — never as literal strings, in either direction. A renamed tool would be invisible to this check. We can report a plausible listing, not a clean one.
  • The vendor's capability list omits an entire group the directory ships. Remesh's "What You Can Do" section names workspace and conversation setup, Remy workflows, and analysis tools. It does
  • One scope covers every write. read and write is a real boundary, and it is only one. It does not fall between the routine write and the irreversible one: granting write grants update_section, delete_message and apply_recruit_audience together. There is no way to let an agent draft a guide without also letting it delete one.
  • No accounting tool exists. Nothing in the surface reads a credit balance, quota or bill, while apply_recruit_audience commits against On Demand Recruit credits. Remesh publishes no per-participant rate, so a recruitment configuration's cost is unpredictable rather than small.
  • We could not read tool schemas or safety annotations. The endpoint returned 401 to an anonymous request on 2026-08-23, so no parameter-level detail is published here — including whether any parameter carries a publish or send flag.
  • Analysis tools require the conversation to have ended. Remesh states its post-conversation analysis suite becomes available only after a conversation ends or external data is imported, and that exporting mid-conversation returns partial data. An agent querying a running study will get less than it would after the fact.
  • Some capabilities are gated on LLM features. Remesh notes several analysis features require a workspace admin to enable Large Language Model features, and that disabling Generative AI disables MCP access by default. A workspace with LLM features off will see a reduced surface.
  • Rate limits apply by workspace and by user. Remesh does not publish the numeric limits, only that both levels exist and that large request batches reach them.
  • Remesh's own setup guide is out of date on verification status. Its MCP setup guide states Remesh "is currently going through the formal verification process" and instructs readers to add a custom connector that will show an unverified warning. Anthropic's directory listed Remesh as a

Frequently asked questions

Yes, and that is the most important fact on this page. Tools including get_sentiment_examples, get_top_responses_by_agreement and get_responses_by_segment_submission return the verbatim text participants wrote. Remesh states its Conversation Data export contains every response to every question asked. Participant answers, not just aggregate statistics, enter your agent's context.

No, and Remesh says so directly. Its recruiting documentation states that Remy sets recruitment up and that you publish and launch from the Build page yourself. No tool among the 52 publishes, launches, ends or moderates. create_conversation_messages writes discussion-guide items into a draft, which is authoring, not sending.

Twenty-one write and two delete. The deletes are delete_message and delete_section. The writes are the create, update, duplicate, set, upload, apply, schedule, run and continue verbs — including apply_recruit_audience and upload_participant_data. A risky-verb scan that looks only for delete would miss nineteen of them.

One tool can. apply_recruit_audience configures Remesh Recruit, and Remesh lists On Demand Recruit credits as a prerequisite for recruiting through Remy. No tool among the 52 reads a credit balance, a quota or a bill, so an agent can arrange a spend with no tool that could check the remaining balance first.

Two, read and write. Remesh's RFC 9728 resource descriptor and its authorization server both advertised scopes_supported of read and write on 2026-08-23. That is a genuine boundary between reading a study and changing one, but it is a single boundary — write covers deleting a section and configuring paid recruitment alike.

No. Remesh states the MCP server exposes only data and actions you already have access to, follows your existing Remesh permissions, and that connecting a client does not grant new permissions. A workspace admin must also enable MCP Access in Manage Workspace settings before any connection works.

A Remesh subscription and an admin who has enabled MCP access. Remesh states an admin must open Manage Workspace, then Settings, find the MCP Access tab and enable the workspace toggle. If Generative AI is disabled for the workspace, MCP access is disabled by default too.

Yes, applied by workspace and by user. Remesh states requests are limited at both levels to protect platform performance, and that large batches of requests may reach those limits. Its guidance when you see a rate-limit message or slow responses is to wait before retrying and send fewer requests at one time.

Sources

Use in Agentman

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

Open in Agentman Studio

Server Info

Category
Analytics
Developer
Remesh
Tools
52
Domain
live.remesh.chat

Using Claude Desktop or another MCP client? Setup docs — the connection URL above works anywhere.