montón

Help

You, your team and your agents write together. People comment on what to change, agents revise and reply, and you keep going until the doc is right. We save every version.

What Monton is

Monton is a shared workspace where people and AI agents write docs together: plans, specs, overviews and notes. People read and edit docs in the web app, and comment on passages. Agents connect through MCP or the monton command, read the comments, revise the docs and reply. We keep every version of every doc, so you can see what changed and bring back an old version. You can download a workspace or a folder as a zip of its markdown files from its menu at any time.

People are at the center: they decide what a doc should say, and agents do the drafting and revising they ask for.

To start, sign up or sign in. Then connect your agent.

Connect your agent

You can connect any MCP client to Monton at https://monton.dev/mcp. The first time you connect a client, you sign in to Monton through it.

Claude Code

  1. Run this command in a terminal.

    claude mcp add --scope user --transport http monton https://monton.dev/mcp

  2. Start Claude Code, type /mcp, pick monton, and sign in to Monton.

Claude apps

Follow these steps in Claude: on the web, in the desktop app, or in the mobile app.

  1. Open Monton in Claude’s directory, then choose Connect.

    Open Monton in Claude’s directory

  2. Sign in to Monton when Claude asks.

If the directory isn’t available, add Monton as a custom connector. On a Team or Enterprise plan, an owner adds Monton first, from the directory or as a custom connector.

  1. Open Settings, then Connectors, and choose Add custom connector.

  2. Name it Monton, and paste this URL.

    https://monton.dev/mcp

  3. Sign in to Monton when Claude asks.

ChatGPT

  1. In ChatGPT’s settings, open Apps (Connectors in older versions), then Advanced settings, and turn on developer mode.

  2. Create an app named Monton with this URL. If ChatGPT asks how to sign in, pick OAuth.

    https://monton.dev/mcp

  3. Sign in to Monton when ChatGPT asks.

  4. In a new chat, open the + menu, choose Developer mode, and pick Monton.

Other MCP clients

For OpenClaw, scripts and other MCP clients that send a token.

  1. Create an agent on the Agents page, and copy its token.

    Open the Agents page

  2. In your client’s settings, add an MCP server with this URL.

    https://monton.dev/mcp

  3. Send the token in this header.

    Authorization: Bearer <your token>

Coding agents, with the CLI

With the monton command, coding agents such as Claude Code or Codex edit a workspace’s docs as files. For comments and the inbox, they also connect Monton as an MCP server, with the steps for their client.

  1. Install it. It needs Node 24 or newer. To try it without installing, run npx @monton/cli wherever these steps say monton.

    npm install -g @monton/cli

  2. Sign in.

    monton login

  3. Copy a workspace to a folder.

    monton clone <workspace>

  4. Ask your agent to read the guide for agents.

    monton help agents

Skills

Add a skill so your agent knows how we work in Monton: it checks the inbox, revises from comments and replies in the thread.

  • Claude Code: unzip the .zip into ~/.claude/skills/, which makes the skill’s folder.
  • claude.ai: open Settings, find Skills, and upload the .zip.
  • Other clients: if yours takes skills, add SKILL.md in its settings.

What each tool does

Agents work in Monton through these MCP tools.

Reading

list_workspaces
Lists the workspaces you can reach, with your role in each.
get_tree
Lists a workspace’s docs and folders, including any AGENTS.md for the agent to read.
read_doc
Reads a doc’s text and version, or an old version of it.
search
Finds docs by their text.
get_history
Lists the changes to a workspace or to one doc.
get_diff
Shows how a doc changed between two versions.

Writing

create_doc
Creates a doc.
create_docs
Creates up to 50 docs in one change.
edit_doc
Replaces exact passages in a doc.
write_doc
Rewrites a whole doc, and merges the new text with what people typed meanwhile.
move_doc
Renames or moves one doc.
delete_doc
Deletes one doc. We keep its text in history.
restore_doc
Brings back a deleted doc, with its history and comments.

Comments and the inbox

list_comments
Lists the comment threads on a workspace or a doc.
comment
Starts a thread on a passage of a doc.
reply
Replies in a thread.
edit_comment
Edits a comment the agent wrote.
resolve_comment
Resolves a thread, or reopens it.
get_inbox
Lists the agent’s mentions, replies in its threads and new threads on its docs.
mark_done
Marks inbox items done.

Imports

start_import
Gives you a link to import a folder from your computer in the web app.
get_import
Shows what an import brought in.

Workspaces and people

create_workspace
Creates a workspace that you own.
list_members
Lists a workspace’s members and their roles.
add_member
Adds a person to a workspace you own, or makes an invite link for them.
share
Shares one folder or doc of a workspace you own with a person.

Prompts to try

Once you’ve connected your agent, ask it:

  • Read Start here in my Monton workspace, then write a one-page plan for something I’m working on, and tell me when it’s ready for comments.
  • Check my Monton inbox. For each comment on my docs, make the change it asks for, reply in its thread to say what you changed, and mark the item done.
  • Search my Monton workspace for what we decided about pricing, and tell me which docs say it.
  • Put the plan from this chat in a new doc in my Monton workspace, in the plans folder.

When a doc is ready, comment on it in Monton, then ask your agent to answer your comments.

Troubleshooting

I sign in and end up back at the start

If you sign in and land back on Monton’s home page, or you see “Sign-in didn’t stick”, check that you allow cookies for monton.dev, and turn off any extension that blocks them for this site. Then sign in again.

If your agent keeps asking you to sign in, sign in to monton.dev in your browser once with the same account, then connect the agent again.

My agent can’t see a workspace

An agent reaches only the workspaces of the person who connected it, as an editor at most. Ask it to call list_workspaces, and check that it lists the workspaces you see in the web app. If it doesn’t, you probably connected it with another account: remove the connector or agent, then connect it again while you’re signed in to the right account.

If someone shared only a folder or a doc with you, your agent sees only that. If you revoked the agent on the Agents page, connect it again.

My agent sees “You’re writing quickly”

Each agent may make 60 changes a minute. When it makes more, we refuse the change, change nothing, and say how many seconds to wait. Ask your agent to wait that long and try again, or to create several docs in one create_docs call.

We marked a comment outdated

We mark a comment outdated when someone deleted or rewrote the passage it was on. We never move it to another passage by guessing, since a comment on the wrong passage is worse than an outdated one. Read the thread, then reply, or resolve it if it no longer applies.

Get help

Email enrique.c.enriquez.m@gmail.com. Tell us what you did, what you expected, and what happened, with any message you saw in Monton.