Connect your AI agent

Back

User guide

Open the AI agent app you use
An agent app is an AI assistant that can do things for you, such as Claude or ChatGPT.
Copy the setup prompt and send it to your agent
Connection prompt for your agent
Connect to Nuto for memory and task tracking. Read the guide: /start.md Use the service environment in the guide without changing this application’s backend address. Set up as much as this host supports: the Nuto MCP connector if the host can add a remote MCP server; the nuto CLI, signed in, if you can run a terminal (with its hooks where the host has them); and the Nuto skill wherever the host loads skills. At least one channel must really read and write. I authorize installing and signing in to the nuto CLI and installing the Nuto skill and hooks here, keeping existing instructions. Sign-ins happen in my browser. After connecting, use the real AGENTS.md, existing agent memory and this conversation that this agent can access. Check duplicates, then prioritize saving valuable user facts, preferences and decisions as Nuto memories; reuse or update existing records. Next, create issues only for real, clear pending work. Use dates only when supported by the real context; a date alone does not justify inventing a time. Without a known date, explicitly leave the issue unscheduled (CLI: --unscheduled; MCP: idea:true), with no automatic scheduling. Do not invent facts, tasks or dates, or upload entire files, secrets or logs. Only if there is no valuable real content to save, save one short greeting memory clearly labelled as an onboarding example. Authorization, deduplication or write failures must not trigger that fallback. Read back saved records to verify them. Never ask for passwords or secrets. Interrupt only when my action is needed: give the required action, authorization link or pairing code, without narrating installation. Finish in one or two sentences with what was actually saved as memories and issues, and ask me to view it in Nuto. On failure, give only the cause and next action; never claim an unsuccessful save succeeded. Keep the current login running while awaiting approval and check its result when the host allows. On resuming, check the existing connection first; with the CLI, run nuto auth whoami before another login. Restart only after authorization expires. Give the CLI's actual approval URL and pairing code (valid for 10 minutes), then wait for me to approve.
Finish what your agent asks
Its last reply says what is left: approve Nuto on the sign-in page that opens, or enter its pairing code (like BCDF-2345) here.
Pairing codes last 10 minutes. If one expires, tell your agent: “The pairing code expired, please generate a new one.”
You can also enter the pairing code in the Nuto mobile app.

Prefer to set it up manually? See manual setup

User guide

Open the AI agent app you use

Copy the setup prompt and send it to your agent

Finish what your agent asks