Open your Alcoven project in Claude

An Alcoven project holds its brief, documents, tasks and notes. People and their AI tools read from it and write to it. Claude, Claude Code and other AI apps all start from one address, the same for everyone:

https://app.alcoven.co/mcp

Add Alcoven to Claude

You add Alcoven on claude.ai or in the Claude desktop app. How you add it depends on your Claude plan.

Free, Pro or Max plan

In Claude on claude.ai or in the desktop app, go to Customize > Connectors.
Click Add custom connector.
Name it Alcoven, and paste the address https://app.alcoven.co/mcp.
Click Add, then Connect.

Or open Claude’s dialog with the name and address filled in, then click Add and Connect. Claude’s Free plan allows one custom connector, and that one covers every project you’re on.

Team or Enterprise plan

An Owner adds Alcoven once, for the whole organization:

Go to Organization settings > Connectors.
Click Add, then Custom. If Claude asks, choose Web.
Name it Alcoven, paste the address https://app.alcoven.co/mcp and click Add.

Owners can also open the dialog with the name and address filled in, then click Add.

Members without the Owner role can’t add a custom connector themselves, so they ask an Owner. Each member finds Alcoven under Customize > Connectors, labeled Custom, and clicks Connect to sign in once with their own Alcoven account. That sign-in reaches every project they’re on.

When you click Connect, Claude opens Alcoven’s sign-in page.

Sign in with your email and password, Google or GitHub. If you don’t have an account, create one free at app.alcoven.co.
Alcoven names the projects you’re on and asks if Claude may work in them as you.
Click Allow.

That one sign-in reaches every project you’re on, and the projects you join later. There is no project to choose when you sign in, and no second connector to add for another project. When you leave a project, Claude stops reaching it.

Use it

Ask Claude about any of your projects by name, in any conversation where the Alcoven connector is turned on. For example:

“What’s waiting for me on the Kiln Coffee launch?”
“Brief me on the pricing task in the Kiln Coffee launch.”
“Does this draft agree with the brief of the Kiln Coffee launch?”

Each request Claude makes is for one project; to work in another, name it. If you’re on more than one project and don’t say which, Alcoven answers with the list of your projects, so Claude can ask you.

You can also start from four ready prompts: Brief me on a task, Ask the project, Check this against the record and What’s waiting for me. If you’re on one project, its own skills appear among them. The AI you open the project from is given its skills too.

What Claude can do there

In your project, Claude has 26 tools: 11 that read and 15 that write.

Tools that read

Open the project (open_project): Its brief and folders, what waits on you (your tasks, the notes that ask you, your next step and why), who looks after what, its skills and schedules, and what changed since you last looked.
Search the project (search): Finds the documents, sections, tasks, notes and files that talk about a subject, a person or a client, ranked, with the address of each.
Find lines in the project’s files (grep): Every line that holds an exact phrase or a pattern, across documents, text files and code.
Read from the project (read): A document, a section, a task or a file; for a schedule, when it runs next and how its last runs went.
List documents (list): What is in a folder or the whole project, newest change first, with who changed each.
What changed (what_changed): What is new since you last looked, or since a day you name, with what waits on you first.
Who wrote it, and its versions (history): Who wrote each passage, with which AI and when, what they asked for and what they did themselves, then every earlier version.
Who sees a folder or file (who_sees): Who can read it and who can change it.
Ask a source for numbers (numbers): How a number is doing over days, weeks or months, from one of the project’s data sources or its own tasks and notes, with where each number came from.
Check a schedule before writing it (check_schedule): Reads a draft schedule the way Alcoven will, and says which lines it can’t read, or when it runs next.
Alcoven’s guidance (guide): The rules and steps for one topic, such as notes, who sees what or schedules, for your AI to follow.

Tools that write

Start a project (create_project): A new project with the name you give it, in your team.
Add a document (add_document): A new document or text file, under your name.
Save a file (add_file): A PDF, a picture, a Word file, slides, a spreadsheet or a ZIP, or a new version of one.
Change documents (edit_document): Up to eight edits to passages, each with its reason. Nothing changes until you say yes.
Confirm what the person said yes to (confirm): Makes a change you saw and said yes to: edits, a skill, a change sent to GitHub, or a change to who sees what.
Add a note (add_note): A comment or a question on a document, a passage or a task, which can ask one person.
Close a note (close_note): Closes a note with what came of it: the document changed, or it stays as it is and why.
Add or close a task (task): A task for you or someone else, with a due day and what it should bring, counted by one of the project’s sources; later, what came of it.
Set someone’s role (set_role): A person’s role on the project and the topics, folders and files they look after.
Change who sees what (set_access): Who can see a folder or a file, which folders one person sees, or who is in a group, once you have seen what it opens or closes, and to whom.
Ask for something closed to the person (ask_for): Sends a request for access to the people who can open it to you.
Send a change to a linked repository (send_change): Sends what changed in a linked folder to GitHub, as a commit or a pull request, once you say yes.
Add rows to a table (add_rows): Adds rows to a table the team keeps, such as a CSV file in the project or a Google Sheet it may add rows to.
Add or update a skill (add_skill): Adds a skill from Alcoven’s skills directory to the project, or updates one, once you have seen it.
Run a schedule now (run_schedule): Runs a schedule’s script once, straight away, once you say yes, and returns its result and the end of its log.

Limits

The work counts toward your Claude plan like any conversation, and costs nothing extra in Alcoven. In each project, Claude can make up to 60 tool calls a minute for you, and up to 20 of them can be Read from the project.

Schedules

A schedule is a file ending in .schedule.md that runs a script of the project (node, python or bash) on a timetable, on Alcoven’s servers, so it runs when nobody’s computer is on. It runs as the person who keeps it: it reads what they can open, and changes only the files it names. A key it needs is named in the file and pasted on the project’s Secrets page by someone who manages the project. Ask Claude to set one up: Check a schedule before writing it gives the file’s format, and Run a schedule now tries it once you say yes. People who write in its folder can also press Run it now in Alcoven.

Changes you confirm

Before Claude changes a document, Alcoven has it show you the edits and their reasons and wait for your yes. Where Claude can show a form, the form is your yes. Otherwise Claude shows the edits first and asks your permission before it uses Confirm what the person said yes to. If you set Claude to use that tool without asking, it will not wait for you.

Adding a skill, changing who sees what, sending a change to a linked GitHub repository and running a schedule now wait for your yes the same way.

New documents, notes and tasks go on the project directly.

Everything Claude writes goes on the project under your name, marked as from Claude, for everyone on the project to see. If you only read a project, Claude can read it and add notes under your name, as you can in Alcoven.

What Alcoven sees

Alcoven sees and keeps the tool calls Claude makes (which tool, and what it asked for, which can include text from your conversation, such as a draft you ask it to check) and what Claude writes. Alcoven never sees the rest of your conversation with Claude. The privacy policy has the rest.

Claude Code and other AI apps

Claude Code

Add Alcoven once on each computer, from your terminal. It then works in every folder:

claude mcp add --scope user --transport http alcoven https://app.alcoven.co/mcp

Then run /mcp inside Claude Code, choose alcoven and sign in. That sign-in reaches every project you’re on. You can also clone a project with git: the project’s Use with git page in Alcoven shows how. A computer is set up once, with one token for every project you’re on, and after that each project is one git clone line. In a project’s folder, its AGENTS.md names the project, so Claude Code works in that one.

Codex

Add Alcoven once on each computer, from your terminal:

codex mcp add alcoven --url https://app.alcoven.co/mcp

Then run codex mcp login alcoven to sign in. As in Claude Code, that sign-in reaches every project you’re on, and in a project’s folder, its AGENTS.md names the project, so Codex works in that one.

GitHub Copilot in VS Code

On Copilot Business or Enterprise, first check with an admin that MCP servers in Copilot is on for your organization. Then, in VS Code, open the Command Palette, run MCP: Add Server, choose HTTP, paste https://app.alcoven.co/mcp and name it alcoven. When VS Code asks where to save it, choose Copilot Global or your user profile, so it works in every folder. The first time VS Code reaches Alcoven, a browser opens for you to sign in, and that sign-in reaches every project you’re on. Ask Copilot Chat about a project by name, in agent mode. Copilot may ask you to confirm each tool it uses.

Cursor

Add Alcoven on Cursor’s Customize page, or put it in ~/.cursor/mcp.json to have it in every folder you open in Cursor. If the file already lists other servers, add only the alcoven entry inside mcpServers:

{
  "mcpServers": {
    "alcoven": {
      "url": "https://app.alcoven.co/mcp"
    }
  }
}

Then sign in when Cursor asks. That sign-in reaches every project you’re on. Cursor asks before it uses each tool, unless you tell it not to.

ChatGPT

You can add Alcoven to ChatGPT on its Business, Enterprise and Edu plans, and only in ChatGPT on the web. On Business, only an admin or owner can add it, once for the whole ChatGPT workspace. On Enterprise and Edu, members an admin allowed can add it too.

In ChatGPT on the web, open Plugins, click +, then Add custom MCP server.
Name it Alcoven, paste the address https://app.alcoven.co/mcp as its URL, and set its authentication to OAuth.
Tick I understand and want to continue, then click Create as a plugin.
Install it, sign in to Alcoven when ChatGPT asks, and click Allow. That sign-in reaches every project you’re on.

On Business, if you’re not an admin or owner, send one of them the address and these steps.

On Pro, OpenAI lets a custom MCP server only read and fetch. You can add Alcoven the same way, and ChatGPT can read your projects but can’t change them. To change them from your AI as well, use Claude’s Free plan, below.

On Free, Go and Plus, ChatGPT can’t add Alcoven. Claude’s Free plan works at no cost: it allows one custom connector. Add Alcoven as that one (the steps above), and it reaches every project you’re on.

From OpenAI’s help page, read on 22 Oct 2026.

Other AI apps

If another AI app you use takes a custom connector (an MCP server), add the same address there and sign in the same way. In Microsoft 365 Copilot, only your company’s admins or developers can add a connector, and most of the ways Microsoft gives them need a client secret, which Alcoven doesn’t give out. If your company wants Alcoven there, write to info@alcoven.co. Gemini in Gmail and Docs takes only the connectors Google lists, such as Asana and Salesforce, and your admins turn those on or off.

What if IT won’t allow MCP connectors in Claude?

On Claude Team and Enterprise, an Owner adds Alcoven once, for the whole organization (the steps above), and each member signs in with their own Alcoven account. So what you ask IT for is one custom connector, added once.

The same project has two other ways in, and what anyone changes through one shows in the others:

The Alcoven app, in your browser at app.alcoven.co, where you read and write the project yourself.
A git folder. Clone the project with git and work in the folder with Claude Code, which reads its CLAUDE.md, or Codex, which reads its AGENTS.md. When you push, your changes go on the project under your name.

Which of these your company allows is for your IT department to say.

Remove it

In Claude, remove Alcoven under Customize > Connectors. On a Team or Enterprise plan, you click Disconnect there, and an Owner removes it for the whole organization under Organization settings > Connectors.

In Alcoven, your account page lists the AI tools you let in, under AI tools. Remove ends a tool’s access to every project at once.

If something goes wrong

“You are not on any project yet”

The account you signed in with is not on any project yet. Start one at app.alcoven.co, or ask someone on a project to invite the email address you signed in with. Then click Connect in Claude again.

Claude can’t find a project

Claude reaches only the projects your account is on. If it asks for another, Alcoven answers with the list of yours. If a project is missing, ask someone on it to invite the email address you signed in with.

“This link cannot be used”

The request from your AI tool was not valid. Remove Alcoven from the tool and add it again. On a Claude Team or Enterprise plan, ask an Owner to do this.

Help

For anything else, write to info@alcoven.co.