Create an agent token in Juscribe, under Settings → Security → Agent Tokens (https://app.juscribe.ai/settings/security). Everything the AI does is credited to that token's agent.
Add the server to your tool's MCP configuration, with the token:

{
  "mcpServers": {
    "juscribe": {
      "command": "npx",
      "args": ["-y", "@juscribe/mcp"],
      "env": { "JUSCRIBE_API_TOKEN": "paste your token here" }
    }
  }
}

Tool	Where that goes
Claude Desktop	Settings → Developer → Edit Config, which opens `claude_desktop_config.json`
Cursor	`~/.cursor/mcp.json`
Gemini CLI	`~/.gemini/settings.json`
VS Code	`.vscode/mcp.json`, with `"servers"` in place of `"mcpServers"` and `"type": "stdio"` added
Codex	`~/.codex/config.toml`, in TOML, below

[mcp_servers.juscribe]
command = "npx"
args = ["-y", "@juscribe/mcp"]
env = { JUSCRIBE_API_TOKEN = "paste your token here" }

Cursor resolves ${env:NAME} in mcp.json, so the token can stay out of a file that might be committed: "JUSCRIBE_API_TOKEN": "${env:JUSCRIBE_API_TOKEN}". Cursor approves the configuration with the variable filled in, so approve the server (cursor-agent mcp enable juscribe) with the variable set as it will be when you run.

It needs Node.js 20 or newer. npx -y fetches the newest release each time your tool starts the server, so there is nothing to upgrade. If your board is not at app.juscribe.ai, set JUSCRIBE_BASE_URL beside the token.

Tools

Tool	What it does
`get_workspace_summary`	The board at a glance: Current and the Backlog, the projects, the iteration
`list_tickets`	Tickets filtered by panel, state, type, project, person, label, date or text
`get_ticket`	One ticket in full, with its comments and attachments
`get_attachment`	A file attached to a ticket: a photo as an image, a PDF or text file as text
`attach_file`	Save a Markdown, text or CSV file you wrote onto a ticket
`create_ticket`	File a ticket; it goes to the bottom of the Backlog unless placed
`create_tickets`	File a list of tasks in one call, in the order given
`update_ticket`	Change a ticket's fields, or send it back to the Icebox or the Backlog
`append_to_description`	Add notes below a ticket's description without changing what is there
`transition_ticket`	Move a ticket forward: start, finish, deliver or cancel it
`add_comment`	Comment on a ticket, with @mentions
`update_comment`	Edit a comment you wrote
`delete_comment`	Delete a comment you wrote
`move_ticket`	Put a ticket directly above or below another
`get_current_iteration`	This iteration's number, dates, points and tickets
`list_subtasks`	A ticket's steps, in order, with their owners
`add_subtask`	Add a step to a ticket
`update_subtask`	Edit a step, tick it done or reopen it
`move_subtask`	Put a step at a place in the list
`delete_subtask`	Take a step off the list for good
`list_blockers`	What a ticket waits on, or what waits on it
`add_blocker`	Record that a ticket waits on another ticket, a project or an outside event
`resolve_blocker`	Mark a blocker resolved
`list_projects`	The board's projects
`get_project`	One project in full, its plan and its tallies
`create_project`	Start a project
`update_project`	Change a project's name, plan, colour or stakeholder
`list_iterations`	Every iteration the board has had
`list_workspaces`	The workspaces the connection reaches, with the ids the other tools take
`list_people`	The people and agents on the board, with the ids a write takes
`list_labels`	The board's labels
`create_label`	A label the board does not have yet, when the person asks
`search`	Search the board by words or by a ticket or project number

How it works with the board

Your AI works a ticket the way a teammate does: it starts it, comments on it and finishes it, under its own agent's name. It delivers only when you ask, unless your board merges Finished into Delivered, where finishing delivers.
Only a person accepts or rejects delivered work. The server offers neither.
Whose move it is, is data: a blocker, a subtask's owner, a state — never a message somebody has to remember to send.