Build Your First Claude Agent in 60 Minutes: The Daily 3 Rs Assistant
Turn Claude into a standing assistant that briefs you every morning — Research, Reports, Reminders — with zero code.
Agent #1 — your daily "3 Rs" assistant: a reusable Claude Project that runs a Research / Reports / Reminders briefing on demand every morning from one short message.
Sign in at claude.ai and confirm you are on a paid plan before you start.
The claude.ai home screen after sign-in, with the sidebar visible showing the Projects entry.
Projects are the workhorse for this lesson, and Projects require a paid plan (Pro or higher). On the free tier you can chat and prompt, but you cannot create a Project. If you do not have a paid plan yet, you can still read every step and do the prompting parts in a normal chat — just know the saved, reusable agent itself is a paid feature. No surprise paywall mid-build.
Check You should now see the main Claude chat screen with a list or sidebar that includes Projects. If you do not see a Projects option anywhere, you are likely on the free tier — upgrade to Pro to continue, or follow along in a normal chat for now.
Pick Claude Sonnet from the model menu near the message box.
The model dropdown open near the message box, with Sonnet highlighted as the selected option.
Claude offers several models from a dropdown: Opus (most capable), Sonnet (the sensible default), and Haiku (fast and cheap). For a daily assistant, leave it on Sonnet — it is the right balance of quality and speed for everyday briefings, and it is what most people should use. Do not overthink this; Sonnet is the correct beginner choice.
Check You should now see Sonnet selected as the active model near where you type. If the menu shows a different name, click it and choose Sonnet.
Create a new Project and name it My 3 Rs Assistant.
A newly created, empty Project titled My 3 Rs Assistant, showing the area where custom instructions are added.
A Project is a reusable workspace. Everything you put into it — a standing brief and any files you add — applies to every chat you start inside it. That is what turns a one-off conversation into a repeatable agent. Find the Projects area, choose to create a new Project, and give it the name above. The name is just a label for you; Claude does not treat the name as an instruction, so the real work happens in the next step.
Check You should now see an empty Project called My 3 Rs Assistant with a place to add custom instructions. If you only created a normal chat, go back and look specifically for the option to create a Project, not a chat.
Open your Project's settings, find the custom-instructions field, and paste the standing brief below.
The Project's custom-instructions field filled with the 3 Rs standing brief, brackets replaced with the learner's real name and timezone.
Custom instructions are the agent's standing brief — the rules Claude follows in every chat inside this Project. This is where your assistant gets its job. Paste the block exactly, then change the bracketed parts (your name, timezone, and what your typical day involves) to fit you.
You are my Daily 3 Rs Assistant. My name is [YOUR NAME] and I am in [YOUR TIMEZONE]. Every morning I will send you my raw inputs for the day. Your job is to turn them into one short, calm briefing with exactly three sections, in this order and no others: 1. RESEARCH — anything I asked you to look into or think through. Give me a tight answer or the 2-3 questions I still need to resolve. If you state a fact I would need to trust, label it clearly as (unverified) and tell me how I could check it. Never invent specifics. If you do not know, say 'I do not know' plainly. 2. REPORTS — a 3-5 line summary of where things stand on what I am working on, based only on what I tell you. Do not assume progress I did not mention. 3. REMINDERS — a short bulleted list of what I should not forget today, in the order you think I should do them. End this section with one single line: 'Top priority today: ____' and fill in your best guess, but mark it as YOUR suggestion, not a decision. Rules: Keep the whole briefing under 200 words. Plain English, no jargon, no hype, no flattery. Do not pad. If my inputs are thin, ask me one clarifying question instead of guessing. If a section has nothing in it, write 'Nothing today' and move on.
Check You should now see your pasted instructions saved in the custom-instructions field, with your name and timezone filled in. If the brackets are still there, edit them — Claude will read [YOUR NAME] literally otherwise.
On paper or in a notes app, jot down today's raw inputs for the three Rs.
Your agent is only as good as what you feed it. Take two minutes and write three rough lists — no formatting needed: one thing you want Claude to RESEARCH or think through, a couple lines on what you are working on for the REPORT, and three or four things you must not forget for REMINDERS. Messy is fine; the agent's job is to tidy it. This step is offline on purpose — the deciding starts with you, in your own words.
Start a new chat inside the Project and send the one-line morning trigger followed by your raw inputs.
A chat inside the My 3 Rs Assistant Project showing the trigger message and Claude's three-section briefing reply with the Top priority line.
Because the chat lives inside the Project, it already knows it is your 3 Rs Assistant — you do not re-explain the job every day. Send the short trigger below, then paste the three rough lists you wrote in Step 5 right under it.
Good morning. Here are today's inputs for my 3 Rs briefing: Research: [what you want me to look into] Reports: [what you're working on] Reminders: [things you can't forget]
Check You should now see a short briefing with exactly three sections — Research, Reports, Reminders — and a single 'Top priority today' line at the end. If you got a long, generic essay instead, the chat probably is not inside the Project; check the chat header shows the Project name, and start the chat from inside the Project.
Read the briefing critically and reply with one correction to teach the agent your taste.
The same chat after the learner sends a correction, showing Claude's revised briefing that reflects the new instruction.
Do not just accept it. If the 'Top priority' guess is wrong, say so: 'Actually the proposal is today's priority, not the email — keep that in mind.' If a section is too long, say 'cut Reports to two lines.' This is how you tune the agent without touching the instructions. When a correction is one you will always want, add it as a new line in the custom-instructions field so it sticks for tomorrow.
Check You should now see Claude adjust the briefing to match your correction in the same chat. If you want that change every day, confirm you also added it to the Project's custom instructions — chat corrections alone do not carry to a brand-new chat tomorrow.
Tomorrow morning, open the Project, start a fresh chat, and send the same trigger — your agent is now a daily habit.
A second day's fresh chat inside the Project, showing the trigger sent and a new three-section briefing returned, proving the agent is reusable.
This is the honest version of 'briefs you each morning': you tap into the Project and send one line, and the standing brief does the rest. It is not a robot that wakes up without you — and that is fine for a first agent. If you later want Claude to run a briefing truly on a schedule, on its own, that exists too: it lives in the Claude desktop app's scheduled tasks (paid, desktop-only, and it leans on your computer being awake — so runs can skip when the machine is asleep). That is the next door to walk through later, not today's lesson. Today, you have a real, reusable agent.
Check You should now be able to get a fresh 3 Rs briefing any morning in under a minute by opening the Project and sending the trigger. If a new chat ignores your rules, confirm you started it from inside the Project, not as a standalone chat.
You shipped it.
Open claude.ai, create a Project called "My 3 Rs Assistant," paste the custom instructions from Step 4, then send the one-line morning trigger from Step 6. When you get a clean briefing back, your agent is shipped.
You just automated the assembling and formatting of your morning briefing — the boring clerical part. You did NOT automate the deciding. Claude can sort your three priorities into a tidy list, but it cannot know which one actually matters today, or whether a "reminder" is worth your stress. Notice the line: the machine arranges, you choose. Before you accept any briefing, read it and ask "is this actually my priority, or just the one that was easiest to phrase?" Keep that judgment muscle yours. Also: where the briefing makes a factual claim you can't verify, make Claude flag it as unverified rather than letting it sound certain — an honest "I don't know" beats a confident guess every time.
Agent #1 — your daily "3 Rs" assistant: a reusable Claude Project that runs a Research / Reports / Reminders briefing on demand every morning from one short message.