Maren MCPRemote MCP · OAuth · 10 research tools

Your AI agent can now run customer research

Connect Claude, ChatGPT, Codex, Cursor, or Gemini CLI to Maren. Plan interviews, launch a study, follow the fieldwork, and bring the synthesis back into the conversation where the work started.

Connect Maren to an agentSetup instructions →

Already use Maren? Sign in and open Settings → AI agents.

Research session
You

Set up interviews to learn why trial users stall before inviting a teammate.

Agent

I drafted a Problem Discovery study with a 30-minute guide. No participants have been added and nothing has been sent. Want to review the questions?

Workspace-scoped · permissioned · auditable

Setup

Connect in two minutes

One URL works everywhere. Sign in with your Maren account, choose a workspace, and review the permissions. No API keys.

MCP server URLhttps://app.maren.so/api/mcp

Claude

claude.ai and Claude Desktop

  1. Open Settings → Connectors and choose Add custom connector.
  2. Paste the Maren URL and select Connect.
  3. Finish the Maren sign-in, pick a workspace, and approve the permissions.

Claude Code

CLI, desktop, and IDE extensions

  1. Add Maren as a remote server, then run /mcp to sign in.
  2. Or install the Maren plugin, which adds a research workflow skill.
Add server
claude mcp add --transport http maren https://app.maren.so/api/mcp
Or install the plugin
claude plugin marketplace add https://github.com/tuhkunen/maren-agent-plugins
claude plugin install maren@maren-agent-plugins

ChatGPT and Codex

ChatGPT connectors, Codex app, Codex CLI

  1. ChatGPT: turn on Developer mode in Settings, then add Maren as a connector with the URL.
  2. Codex app: Settings → MCP servers → Add server → Streamable HTTP, paste the URL, then Authenticate.
  3. Codex CLI: add the server to config.toml and sign in.
config.toml
[mcp_servers.maren]
url = "https://app.maren.so/api/mcp"
auth = "oauth"
default_tools_approval_mode = "writes"
enabled = true
Sign in
codex mcp login maren
Or install the plugin
codex plugin marketplace add tuhkunen/maren-agent-plugins

Cursor

One-click install or mcp.json

  1. Use the Add to Cursor button, then approve the OAuth sign-in when Cursor prompts.
  2. Or paste the server into .cursor/mcp.json.
Add to Cursor
mcp.json
{
  "mcpServers": {
    "maren": { "url": "https://app.maren.so/api/mcp" }
  }
}

Gemini CLI

Extension with the MCP server bundled

  1. Install the extension, then authenticate with /mcp auth maren.
Install
gemini extensions install https://github.com/tuhkunen/maren-agent-plugins

Any other MCP client

Streamable HTTP with OAuth 2.1

  1. Add the URL as a remote MCP server. Maren answers unauthenticated calls with a 401 and discovery metadata, so clients that support OAuth with PKCE and dynamic client registration sign in without an API key.
Server URL
https://app.maren.so/api/mcp

Team and Enterprise plans on Claude: an organization owner adds the connector first, then each member signs in to Maren and approves a workspace. Disconnect any client from Maren Settings → AI agents.

An operating layer for research

From “we should talk to users” to a live study

Your agent already knows the product question, the decision behind it, and the context scattered across your conversation. Maren MCP gives it a safe way to turn that context into research work instead of another document you have to translate by hand.

  1. 01

    Plan with live context

    The agent can check workspace limits, find existing projects, and compare Maren’s interview methods before it proposes a study.

  2. 02

    Build a research project

    Turn the goal already in your conversation into a draft project with a chosen style, depth, language, tone, and discussion guide.

  3. 03

    Prepare participant links

    Add participants, activate the project, and retrieve a unique interview link for each person. Maren does not let the agent email anyone.

  4. 04

    Follow the fieldwork

    Ask for a compact status snapshot instead of opening another dashboard: pending participants, completed interviews, and the next useful action.

  5. 05

    Bring findings back

    Start cross-interview synthesis when enough interviews are complete, then retrieve the finished report in the conversation where the work began.

What changes for customers

Less research administration. More continuity

A good agent should not replace customer evidence. It should make collecting that evidence easier. Maren keeps the interviewer, research methods, participant experience, and synthesis consistent while your agent handles the operational handoffs around them.

01

Stay in the work

Move from a product discussion to a prepared research project without rebuilding the context in another interface.

02

Repeat the good parts

Give the agent a repeatable workflow while Maren applies the same research methods and product rules every time.

03

Keep judgment human

Review the study, decide who to invite, approve sensitive access, and interpret findings in the context of the business.

The complete first-release surface

Ten tools. One research workflow

Each tool has a narrow job, structured inputs, and structured results. That makes the workflow legible to an agent and keeps Maren authoritative for research rules, access, and billing limits.

  1. 01Inspect workspace and quota
  2. 02List interview styles
  3. 03Find existing projects
  4. 04Create a draft project
  5. 05Add participants
  6. 06Activate a project
  7. 07Get participant links
  8. 08Check project status
  9. 09Start synthesis
  10. 10Retrieve synthesis

Try it

Prompts that put Maren to work

Once Maren is connected, talk about the decision you need to make. The agent chooses the tools. These prompts walk through the whole workflow, from checking capacity to reading the synthesis.

  1. “Check my Maren workspace and tell me how many interviews I can still run this month.”

    Workspace and quota

  2. “We keep debating why trial users never invite a teammate. Compare Maren’s interview styles and draft a 30-minute discovery study about that moment.”

    Styles and project creation

  3. “Add these five customers to the onboarding study as participants, activate it, and give me each person’s interview link so I can email them myself.”

    Participants, activation, links

  4. “How is the onboarding study going? Who has finished, who is pending, and what should I do next?”

    Project status

  5. “Four interviews are complete. Start the synthesis, and when it is ready summarise the top three findings with the evidence behind each.”

    Synthesis

Useful does not mean unbounded

Designed around clear edges

Research touches customer identities, private workspaces, and decisions that can cost money. Maren MCP makes those boundaries visible instead of asking you to trust a broad API key.

One workspace at a time

You choose the Maren workspace during connection. Every request is resolved inside that workspace, and membership is checked again each time.

Permissioned by scope

Read projects, create studies, manage participants, access links, and run synthesis are separate permissions you review before connecting.

No silent outreach

Adding a participant never sends an invitation, reminder, or thank-you message. You keep control of who gets contacted and when.

Traceable operations

Maren validates inputs, enforces quotas and rate limits, and attaches an audit reference to every dispatched tool operation.

Questions

Before you connect

Maren uses OAuth, so there is no API key to copy into a chat. Choose a workspace, review access, and disconnect the client from Maren whenever you want.

Open Maren settings →
What is Maren MCP?

Maren MCP is a remote Model Context Protocol server that gives an approved AI agent a structured set of tools for running customer research in Maren.

Which AI agents can connect to Maren?

Claude, Claude Code, ChatGPT, Codex, Cursor, and Gemini CLI have guided steps on this page. Any client that supports remote MCP over Streamable HTTP with OAuth 2.1, PKCE, and dynamic client registration can connect with the same URL.

Can an agent email participants through Maren MCP?

No. An agent can add participants and retrieve their unique links when it has permission, but MCP-created projects do not send invitations, reminders, or thank-you emails.

Does the agent get access to every Maren workspace?

No. You choose one workspace when you connect, review the requested permissions, and can revoke the connection from Maren settings.

Does Maren MCP expose raw interview transcripts?

The first release returns project status and cross-interview synthesis, not raw transcripts. Completed synthesis is explicitly marked as untrusted research data so an agent treats participant content as evidence, not instructions.

Tell Maren what you want to learn

Try Pro for 30 days. Your first interview can be live in five minutes. No credit card required.

Try Pro for 30 days