Guide
Claude only knows what is in the conversation, so give it three things: a short standing brief about your role and projects, a dated log of what you did, and real material such as the actual email or doc. A paper or text log works for most people. Dayflow helps on a Mac when you forget to log, because it builds a timeline from your screen automatically.
You open a chat with Claude, type “help me write my update for Friday,” and get back three tidy paragraphs about a project that sounds a bit like yours. The milestones are plausible, the tone is fine, and none of it is true. So you start explaining. Which client. What actually shipped. Why the launch moved. Twenty minutes later you have typed most of the update yourself, just in a messier order.
Or you ask for help planning next week, and the plan is sensible for a generic person with your job title. It does not know you lost Tuesday to a billing bug, that the design review is the real deadline, or that you already tried the obvious fix and it broke something else.
This guide covers why that happens and how to hand Claude the context that lives in your head and your week. Most of it takes paper, a text file and a few minutes. Dayflow publishes this blog, and our Mac app comes up once near the end, along with where it is the wrong choice.
Claude starts from what is in the conversation. It cannot see your inbox, your calendar, your files or your screen unless you connect something or paste it in. Depending on how you use it, it may carry some notes between chats, but it was not there for your Tuesday. When it lacks a detail, it fills the gap with the most typical version of your situation. That is where the generic answer comes from.
The harder part is on our side. Context is mostly invisible to the person who has it. You know who “Sam” is, what “the migration” means and why the Henderson account is touchy, so those words feel self-explanatory. You leave them out without noticing.
Memory makes it worse. Many people find that by Thursday, Monday is a blur, and what survives is whatever was recent or stressful. Ask yourself what you did this week and you get the dramatic meeting and the last thing you touched, not the four quiet hours of cleanup that were the actual work.
Then there is compression. When you describe your work, you summarize it, and summarizing means deciding what matters. Often the detail you dropped is the one that would have changed the answer: the constraint, the earlier failed attempt, the person who has to approve it.
Dumping everything is not the fix either. A long, unsorted paste buries the point, and the answer tends to drift toward whatever took up the most space. What helps is a small amount of the right context, gathered before you need it.
What to do. Open a plain text file or a page in a notebook and write one page, no more, covering:
Put a date at the top. Paste it in at the start of any conversation about your work.
Illustration. “I’m a product designer on a four-person team. Current work: (1) redesigning the checkout flow, due for review Oct 21, approver is Priya; (2) cleaning up the icon library, no deadline. ‘The audit’ means the accessibility review from August. Write like a calm colleague, short sentences, no hype.”
After a week. Count how often you had to correct a basic fact in Claude’s first answer. If that number dropped, the brief is working. If you keep adding the same missing detail mid-conversation, it belongs in the brief.
What to do. Write two or three lines at natural breaks, such as before lunch and before you close the laptop. Date each entry. Note what you worked on, what changed, and anything that is stuck. Paper works. So does a note on your phone. If you want a structure, the daily work log template gives you columns to fill, and there is a longer guide on keeping a work journal.
The point is to capture things while they still have detail. “Fixed checkout bug” written on Friday is useless. “Found the checkout bug was a rounding issue in tax, patched it, need QA to confirm on Safari” written on Tuesday is exactly what Claude needs.
Illustration. “Tue Oct 7, 12:30. Spent the morning on checkout tax rounding. Patched, waiting on QA. Priya asked for two layout options by Thursday. Haven’t started.”
After a week. Try asking Claude for a weekly summary using only your log. If the draft needs light edits rather than a rewrite, the log is detailed enough. If it reads vague, your entries are probably too short, so add one more line about what changed or what is blocked.
What to do. Before a hard request, say the situation out loud to someone who does not work with you: a partner, a friend, a voice memo to yourself if no one is around. Notice where they stop you to ask “wait, who is that?” or “why does that matter?” Those questions mark the context you skip without realizing.
Illustration. You start with “I need to push back on the timeline.” Your friend asks, “Whose timeline?” Then, “Why is it slipping?” Then, “What happens if it slips?” By the third answer you have the request you should have typed: who it is for, what caused the delay, and what is at stake.
After a week. Look at the first message you typed in your last few conversations. If it now includes the who, why and stakes up front, you have absorbed the habit and can drop the rehearsal for routine requests.
What to do. When the work involves a specific email, doc, message thread or error, paste the relevant excerpt instead of paraphrasing it. Your paraphrase already contains your interpretation. The original lets Claude notice what you missed.
Before pasting, check your employer’s rules about work material in a cloud assistant, and strip out client names, personal details and anything confidential. Keep it short: the paragraph that matters, not the whole thread.
Illustration. Instead of “the client seems unhappy about the delay,” paste their two sentences: “We were expecting this last week. Can you tell me what changed?” Claude can now see that they asked a direct question, which calls for a direct answer, not an apology letter.
After a week. Compare drafts where you pasted the source with drafts where you summarized. If the pasted ones needed fewer rounds of back and forth, keep doing it. If you find yourself pasting whole threads, practice trimming to the one excerpt that matters.
What to do. Start each request with three things: the task, the audience, and what a good result looks like. Add any hard limits, such as length, format or a word your manager hates. This is the part of context people most often skip, because it feels obvious to them.
Illustration. “Task: a status update for my manager. Audience: she reads it on her phone between meetings. Good looks like: under 120 words, the decision I need from her goes first, then two lines on progress. Material below.” If you write status reports often, the weekly status report template gives you a fixed shape to point Claude at.
After a week. Notice how many follow-up messages you sent saying “shorter” or “more casual.” If those dropped, the framing is doing its job. If one particular correction keeps coming up, move it into your standing brief.
What to do. For anything bigger than a quick draft, end your message with: “Before you answer, ask me up to five questions about anything you’d need to know.” Answer them briefly, then let it write. This turns the gap-finding from step three into part of the conversation.
Illustration. You ask for help planning a project kickoff. Claude asks who has final say, what the budget is, whether there is a hard launch date, what went wrong last time, and who is skeptical. Two of those you would never have thought to mention.
After a week. Look at the questions Claude asked. If the same ones keep coming up (“who approves this?”), add the answers to your brief. When the questions start being about genuinely new details, your brief has caught up with your work.
Everything above depends on you writing things down. On busy days that is exactly what slips, and you are back to reconstructing the week from memory, which is the subject of our page on where your day went.
Dayflow is a free, MIT-licensed Mac app for macOS 14 or later that records your screen at one frame every 10 seconds and has AI write the day into a readable timeline. There is no timer to remember. Recordings and the timeline are stored on your Mac. You choose how screenshots are analyzed: Local AI runs a supported model on your Mac through Ollama or LM Studio, so analysis stays on-device; or your own cloud provider or Dayflow Pro ($20 a month) processes selected screenshots off-device. The details are on the local and cloud processing page.
For Claude, the useful part is getting that record out. You can export a date range as Markdown from Settings → Export and paste it in, and the integrations page covers other ways an assistant can retrieve Dayflow data. The Claude weekly report workflow walks through one example. One boundary matters: an assistant receives whatever Dayflow data it retrieves in your conversation, and local storage does not make a cloud assistant conversation local.
These Dayflow details come from its own documentation as reviewed on 2026-09-06. Dayflow is our product, and Dayflow publishes this article.
Your important work happens off the screen. Dayflow sees only what is on your Mac’s screen. Meetings in a room, phone calls, whiteboard sessions and thinking on a walk do not show up. If that is most of your week, a written log beats it, because it captures what happened and why.
You are not on a Mac. Dayflow is Mac only, macOS 14 or later. On any other computer, a paper log and a standing brief are the way to go.
You need the reasons behind the work. A screen record shows what you had open. It does not know you chose option B because the client hates modals. That reasoning lives in your notes, and Claude needs it more than it needs a list of apps.
You want everything to stay local. Local AI keeps analysis on your Mac, but once you paste an export into a cloud assistant, it is in that conversation. If that is not acceptable for your work, trim the export first or keep it out of the chat.
You need billing, approvals or blocking. Dayflow has no invoicing, payroll approvals, team reports or website blocking. If you need hours you can bill, a dedicated time tracker for Mac with invoicing is the better fit; our roundup of Mac time trackers covers those.
You want someone to keep you on track. Dayflow is a record. It is not a coach, a planner or a focus tool. Its AI summaries can also miss or misread work, so check anything before you paste it somewhere that matters.
If your notes are already good, you may not need it at all. A one-page brief and three lines a day cover most of what Claude needs.
Claude can only work with what you give it, and the most useful context is the kind that feels too obvious to mention: who is involved, what already happened, what good looks like. Write a one-page brief and keep it current. Log a few lines while the day is still fresh. Paste the real material instead of your memory of it. Start each request with the task and the audience, and let Claude ask questions before it answers.
Try the brief and the log for one week. If the first drafts get closer and the follow-up corrections shrink, that is the whole system. If the log keeps falling apart on busy days, that is the point where an automatic record of your day might be worth a look.
Claude works from what is in the conversation. Unless you connect a source or paste something in, it cannot see your calendar, inbox, files or screen. When it gives a generic answer, it is usually filling gaps with a typical version of your job. The fix is to bring the specifics yourself: who is involved, what happened, what you already tried, and what a good result looks like.
Keep it to one page. Include your role, the two to four projects that matter right now, the people you work with and how they relate to you, any terms or acronyms that only make sense inside your team, and how you like things written. Date it, and update it when a project starts or ends. A brief you actually maintain beats a long one you wrote once and forgot.
More is not always better. A long, unsorted paste can bury the one detail that matters, and the answer drifts toward whatever took up the most space. Lead with the task and the goal, then add the background that changes the answer. If you are pasting a week of notes, say which parts are relevant. When in doubt, ask Claude what else it would need.
That depends on your employer's rules and on the assistant's own terms, so check both first. Many people remove client names, personal details and anything confidential before pasting. Keep in mind that storing notes locally does not make a conversation with a cloud assistant local. Whatever you paste or let it retrieve is sent to that service as part of the chat.
Memory reconstructs, and by Friday the early week tends to blur. The most reliable approach is to write a few lines as you go, such as a dated note at lunch and at the end of the day. Your calendar and sent folder help fill gaps. If you often forget to write anything, an automatic record of your screen time is another option on a Mac.
Dayflow can export a date range as Markdown from Settings, which you can paste into Claude or attach to a conversation. Its integrations page covers other ways an assistant can retrieve Dayflow data. An assistant only receives the Dayflow data it retrieves in your conversation, and that data then goes to the assistant's service. Check the summary before sharing it, since AI summaries can miss or misread work.
No. Most of what helps is free: a one-page brief in a text file, a paper log, your calendar, and the habit of pasting the real email instead of describing it. An app becomes useful when the bottleneck is remembering to write things down at all. If your notes are already good, you probably do not need anything else.
Not always. A written log wins when your important work happens off the screen, in meetings, on calls, at a whiteboard or in your head, because Dayflow only sees what is on your Mac's screen. A log also captures why you made a decision, which a screen record cannot. Many people get the most out of a short written log plus whatever record they already have.
Partly. Dayflow stores recordings and its timeline on your Mac, and its Local AI mode keeps screenshot analysis on-device through Ollama or LM Studio. But the moment you paste that history into a cloud assistant, it leaves your Mac. If you want everything to stay local, review and trim the export first, and share only the parts the task needs.
Dayflow is $0 and open source. Full recordings stay on your Mac, with local AI available for on-device analysis.
Download for MacOS macOS 14+ · MIT licensed · no account required