Claude for project managers: Projects, status briefs and the decisions that have to stick
If you run client or stakeholder projects, Claude is useful for three things: a Project per engagement, a status brief that keeps decisions apart from assumptions, and a decision log that survives the next change request. Here is how project leads use it, what Anthropic's own documentation says each feature does, one worked example, and the one record a Project never holds: what each stakeholder said, across threads.
Short answer: make one Claude Project per engagement, write project instructions that force every brief into three buckets (confirmed decisions, working assumptions, open questions), and keep the dated records in the Project's knowledge rather than trusting memory. Claude drafts well from what you give it; the status, the dates and the sending stay with you. This page is written for the project or program lead running client work, which makes it a cousin of our map of where these tools help a consulting engagement; its ChatGPT twin covers the same job with OpenAI's tools.
A disclosure. This page is published by Equerry, which makes a personal assistant for iPhone; it holds the stakeholder record the last section describes and does none of the drafting Claude does. Equerry is not affiliated with Anthropic. Every Anthropic statement below was read at the source on 27 October 2026, and the project in the worked example is invented.
How project leads use Claude today
Search for this phrase and page 1 is a mix: write-ups about Claude Projects, a university how-to, a Reddit thread sharing one person's project system, a Udemy course, and, under the "project management" phrasing, tutorials for Claude Code and Cowork. Sorting them by who they are for saves an hour.
- The Projects write-ups are the ones for you. Fifty One Degrees, a UK consultancy, published "Claude Projects: The AI Project Management Revolution" (Nick Harding, 3 December 2024), recommending a Project per initiative and listing "Client engagement, from proposals to delivery" among its uses. UC Today's "Anthropic Projects: Using Claude AI for Project Management" (Rebekah Carter, 15 November 2024) walks through the same idea for research, content and budget work. Both predate Claude's current memory and the Cowork merge, so take the mechanics from Anthropic's pages below, not from them.
- The university how-to is Syracuse University ITS's "How to Use Claude to Create and Manage a Project: Step-by-Step Guide for New Users" (21 November 2025). It is written for the university's Claude Enterprise users, so its steps "Set Project Privacy," "Share Projects" and "Access and Use Shared Projects" assume a Team or Enterprise plan; the steps "Create a New Project," "Add Instructions/Prompts" and "Upload Files" apply to anyone.
- Anthropic's own use case is the most telling. Its page "Generate project status reports" (academy.claude.com, undated, read 27 October 2026) promises to "Pull status updates from your emails, Slack channels, meeting notes, and project tools" through Gmail, Google Drive, Google Calendar and Slack connectors and produce "an Excel tracker with visual indicators and context in cell comments." That is a tracker built from connected accounts, for someone willing to connect them.
- The Claude Code and Cowork results are for a different reader. Claude Code is Anthropic's tool for engineers working in a codebase. Cowork was the name of Claude's agentic desktop mode; since 16 September 2026 Anthropic has been folding it into Claude itself (detail below). If you manage a software team, those tutorials may matter; if you manage a client's office move, a product launch or a regulatory program, the ordinary Claude app with a Project is the tool.
Across the write-ups, three patterns recur: a Project per engagement, a status brief drafted from supplied records, and a decision log maintained inside the Project. The rest of this page does each one against Anthropic's documentation.
What Claude's own documentation says Projects, memory and scheduled tasks do
Anthropic's own pages settle what each feature does. Labels below are quoted from them as read on 27 October 2026; menus change, so trust your own screen where it differs.
Projects
Anthropic's help center defines them plainly: "Projects allow you to create self-contained workspaces with their own chat histories and knowledge bases." The setup path in "How can I create and manage projects?" is: open "Projects," select "+ New Project," enter a name and description (the article's note: "Claude will not have access to these details," so the client's name belongs in the instructions, not only the title), then click "+" to upload documents to the project knowledge base and "Set project instructions" followed by "Save instructions." Projects are "available to all users, including those with free Claude accounts"; "Free users can create a maximum of five projects." On paid plans, when knowledge nears the context limit Claude switches to a retrieval mode that expands capacity "by up to 10x while maintaining response quality." Sharing a Project is limited to Team and Enterprise plans. A Project can be archived from the three-dot menu, and "There is no way to delete archived projects at this time (you must unarchive it first)," so a finished engagement sits in the archive until you unarchive and delete it.
Memory, including a Project's own memory
Claude's memory is not the file you uploaded; it is what Claude noted while you chatted. Anthropic's article "Use Claude's chat search and memory to build on previous context" says Claude "saves memory as a set of individual topics as you chat, rather than summarizing conversations after they end." To see it: "Settings > Your Claude," then "Manage" next to "Your memory." "Everything Claude remembers is listed under Topics"; you "select any topic to read it, then use the edit icon to change it or select 'Delete' to remove it." Two lines matter for project work. "Each project has its own memory," and memory from other projects is not used inside it; "to keep a project's memory fully separate, turn off 'Use account memory' under Memory on the project's page," which keeps one client's memory from colouring another's brief. And deleting a chat does not remove the memories it generated; they are deleted one by one, or all at once with "Reset memory," which "permanently deletes all memories including project memories" and "cannot be undone." Memory is on by default for Free, Pro and Max; "On Team and Enterprise plans, owners control whether memory is available, and it's off for each member until they turn it on."
Scheduled tasks
Anthropic's article "Schedule recurring tasks in Claude Cowork" opens: "Scheduled tasks allow you to delegate work to Claude Cowork." You describe a task once and Claude runs it on a schedule or on demand, producing "reports, briefings, and summaries." They are "available on all paid plans (Pro, Max, Team, and Enterprise)." Each run is its own session and you "can review the results when they're ready"; the article does not describe a task sending anything to another person. For a project lead the natural use is a Monday-morning brief of a Project's own material; the setup path, the frequency options and how to check a run are in our guide to Claude scheduled tasks.
One Claude, and what the plans cost
On 16 September 2026 Anthropic wrote: "Starting today, Claude Cowork and chat are merging into one Claude," and "It's rolling out on Pro and Max plans over the next few weeks, with more plans to follow"; its 23 October update says the unified experience is generally available for Enterprise. The matching help article promises "Your chats, tasks, projects, and settings come with you when your account moves over," that "Projects work in every conversation," and that "Scheduled tasks can run in the cloud, so they keep going when your computer is off." From claude.com/pricing, read 27 October 2026 (US, before tax): Free is $0 with Projects "Up to 5"; Pro is "$20 if billed monthly" or $17 a month "with annual subscription discount ($200 billed up front)," listing "Hand off and schedule tasks" and "Projects"; Max is "From $100" a month; a Team standard seat is $25 a month billed monthly or $20 billed annually. Memory shows "Yes" on every plan. Anthropic's caveat: "Price and plans are subject to change at Anthropic's discretion."
A worked example: the Linden move
Linden is an invented client, a 60-person accountancy practice moving to new offices, and you are the project lead the firm hired to run the move. Three records make up the whole example. Nothing here is a real client or a transcript of a product test.
| Record | What it says |
|---|---|
| A · Kickoff note 2 October | Move weekend agreed with the client: 22–23 November. IT cutover is owned by the client's IT contractor. Furniture order placed with the supplier. The managing partner wants the boardroom usable for the partners' meeting on 1 December. |
| B · Email from the client's operations director 14 October | "Could we bring the move forward two weeks, to 8–9 November? The partners would like a settled office well before the 1 December meeting, and the new lease allows it. Separately, is there any way to end the old lease early?" |
| C · Steering group minute 17 October | Earlier date declined: the supplier has confirmed furniture delivery for 18 November, so 8–9 November is not possible. Move stays 22–23 November; the operations director agreed. Action: project lead to confirm with the IT contractor that cutover finishes by Friday 28 November. Early lease termination to be explored; no decision, no owner recorded. |
Three places a brief goes wrong. The request in B was declined in C, so a brief that lists "client has asked to bring the move forward" as an open item is reporting a settled matter as live. The boardroom being ready for 1 December is an assumption until the IT contractor confirms cutover; a brief that states it as fact has promised something nobody has confirmed. And the early lease termination has no owner in any record; a brief that assigns it to someone has invented a commitment.
Put the rule in the project instructions, not in each prompt
The Claude-specific move is to write the shape of a brief once, as project instructions, so every chat in the Linden Project inherits it. Upload A, B and C to the Project's knowledge, then set instructions along these lines:
This Project is the Linden office move. Every status brief uses three sections in this order: Confirmed decisions (each with the record letter and date), Working assumptions (what we are treating as true, and what would confirm it) and Open questions (the owner and date only if a record names them; otherwise write "no owner recorded"). A request is not a decision until a record shows it accepted or declined. A date is a target until a record confirms the dependency behind it. Never add a status colour. Where records conflict, the later record wins and the earlier one is shown as superseded. Keep client-facing drafts under 200 words.
Then the prompt in each chat can be short: "Status brief for Linden as of 20 October." Here is what a correct brief contains; it was derived by reading A, B and C with a pen, not by running anything. Confirmed decisions: move weekend 22–23 November (A, confirmed C, 17 October); the 8–9 November request (B, 14 October) declined (C); furniture delivery 18 November (C). Working assumptions: boardroom ready for the 1 December partners' meeting, which depends on IT cutover completing by 28 November and is not yet confirmed. Open questions: IT contractor's cutover confirmation, owner the project lead (C); early lease termination, no owner recorded, no date. If a draft puts the move-forward request under open questions, point it at C. If it names an owner for the lease, ask which record says so; if it cannot, blank the owner.
Keep the decision log as a file, and let memory be a convenience
A decision log in this setup is a file in the Project's knowledge that you update, one row per question the project has had to answer: what was asked, by whom and when; what was decided, by whom and when; which record proves it; and what, if anything, the decision still depends on. The trap in project work is the overwrite: the operations director's request replaces yesterday's decision in someone's head. A row that reads "8–9 November requested 14 October; declined 17 October (C); furniture delivery 18 November" stops the next meeting from reopening it. Ask Claude to add the new record's row and to mark any earlier row it overtakes, then save the result back into knowledge. Claude's memory may note that "the client wanted an earlier move date"; that is useful colour for a conversation, but it is not a record anyone can be held to, and it survives a deleted chat. The file is the record.
What never goes into the Project
Client financials, personal data about the client's staff and anything a contract or your engagement letter restricts stay out, however convenient a Project makes uploading them; what Anthropic says it keeps, trains on and lets you delete is set out in our guide to Claude data privacy, and your client's own policy decides the rest.
The record a Project never holds: what each stakeholder said
Look at where the Linden records came from. A is a note you wrote. B is an email in your mailbox. C is a minute someone typed up two days after the meeting, and the moment that decided it, the supplier's "18 November, confirmed," was a phone call and a one-line email a week earlier. The Project knows what you uploaded. Anthropic's route to the rest is connectors: its status-report use case reaches "your email threads, Drive file updates, calendar events, and Slack conversations" by connecting Gmail, Drive, Calendar and Slack, a whole account at a time. That is a reasonable choice for your own employer's systems. For a client's correspondence, a vendor's confirmations and a contractor's promises, spread across six engagements, it is a lot of reach for one question: what did this person say, and when?
Where Equerry fits
Equerry is a personal assistant on your iPhone who helps you get things done. For a project lead that means the stakeholder record, not the plan: it keeps no board, assigns no tasks and drafts no Gantt chart. It reads only what you send it, one thing at a time, and never asks for your inbox.
| What you send | The Linden move thread with the operations director, by adding Equerry's own email address to the CC line; replies that keep it on the thread keep arriving. The supplier's delivery confirmation, forwarded. A voice note you record in Equerry in the car park after the steering meeting: "Told the IT contractor I'd have a decision on the cutover window by Thursday." The signed minute, shared as a PDF from your iPhone's share sheet. |
|---|---|
| What you ask | "What did Linden's operations director ask about the move date, and what did we decide?" "What did I promise the IT contractor after the steering meeting?" "Which version of the move plan did we last send Linden, and when?" |
| What you get back | The answer, in plain words, from what you sent, each item on its own date: the 14 October request, the 17 October decision that declined it, your Thursday promise from the voice note, and the plan version from the thread you CC'd, if that email carried it. When an answer draws on a note you saved, it carries a source chip: tap it to see the passage Equerry used and open the original. For a thread you CC'd or forwarded: What Equerry remembered from it is listed in Settings → Your memory, each line marked as from that email, and you can delete any line. |
The difference from a connector is the unit. A connector gives Claude an account; Equerry takes the one thread you CC, so what it holds about Linden is exactly what you chose to send, and the same is true of the other five engagements. If the steering call is on your calendar, ask five minutes before it; if you want the nudge, set a Reminder for that morning with the text you want to see.
The honest limits. Equerry does not maintain a project board, assign work, hold a shared team memory, sync with your project system or send status emails; you write the update, and you press send. It cannot find the thread you never forwarded, and anyone on a CC'd thread can see its address on the CC line, so forward instead where that would be awkward. Sending something is still sharing it: what you send is processed by the outside providers named on our AI disclosure page, so your client's policy applies here exactly as it does to a Claude Project. Equerry Pro is $39.99 a month or $329 a year.
Claude's Project drafts the brief from the records you uploaded; Equerry keeps what each stakeholder said so the records exist when you need them. Before the status call, that is the question that matters: not what the plan says, but what was decided, what you promised and what went out.
Frequently asked questions
Can Claude manage a project?
It can hold one project's material in a Project, draft a status brief from records you give it, keep a decision log you maintain as project knowledge and, on a paid plan, run a scheduled task that produces a report or briefing. Anthropic's own project-management use case (academy.claude.com, read 27 October 2026) is titled "Generate project status reports" and describes pulling updates "from your emails, Slack channels, meeting notes, and project tools" into "an Excel tracker." What it does not do is own the plan: nobody is chased, no board is updated and nothing is sent to a stakeholder unless you send it. Treat it as a very fast analyst who drafts; the status, the dates and the sending stay yours.
What is a Claude Project, and how is it different from a chat?
Anthropic's help center ("What are projects?", read 27 October 2026) describes Projects as "self-contained workspaces with their own chat histories and knowledge bases." A chat is one conversation; a Project is a container for many, with files you upload as project knowledge and project instructions that shape every chat inside it. Each Project also has its own memory; Anthropic's memory article (read the same day) puts it this way: "While working in a project, Claude uses the project's memory alongside your memory from chats outside of projects, but not memory from other projects." Projects are "available to all users, including those with free Claude accounts"; Free accounts can create a maximum of five.
Can Claude send my status report?
Anthropic's scheduled-tasks article (read 27 October 2026) says scheduled tasks are "available on all paid plans (Pro, Max, Team, and Enterprise)" and describes outputs such as reports, briefings and summaries, each run being its own session where you "can review the results when they're ready." The article does not describe a task emailing a client or stakeholder, so plan on reading the draft and sending it yourself. For a client update that is the right boundary anyway: a status report is a commitment the moment it leaves you, and a schedule should not make it.
Does Claude remember decisions across a project?
Two different things carry across chats. Project knowledge is whatever you uploaded, and it stays exactly as you left it. Memory is different: Anthropic's help center (read 27 October 2026) says Claude "saves memory as a set of individual topics as you chat, rather than summarizing conversations after they end," that "each project has its own memory," and that when a chat is deleted the memories it generated are not removed. For a decision that has to survive a dispute, do not rely on memory; put the dated decision record into the Project's knowledge and ask Claude to cite it. The memory topics are a convenience; the file is the record. More on the mechanism in our guide to whether Claude has memory.
Should a project manager use Claude Code or Cowork instead?
Probably neither by name. Claude Code is Anthropic's tool for software engineers working in a terminal and codebase; a project lead who is not shipping code has little reason to open it, even though it shows up on page 1 for "claude for project management." Cowork was Anthropic's name for Claude's agentic desktop mode, and on 16 September 2026 Anthropic wrote: "Starting today, Claude Cowork and chat are merging into one Claude," rolling out "on Pro and Max plans over the next few weeks, with more plans to follow." If your account already has the new experience you will not see separate Chat and Cowork options; you simply ask Claude, and Projects "work in every conversation."
Where does Equerry fit next to Claude for a project manager?
Claude's Project knows what you uploaded to it. The stakeholder record of a live project lives somewhere else: in the email thread where the client asked for a change, the one where you declined it, the supplier's confirmation and the note you made walking out of the steering meeting. Equerry is a personal assistant on your iPhone for that record. You CC its address on the one thread, forward the confirmation, record a voice note after the call, and before the next status call you ask what was decided, what you promised and which version went out. It keeps no board, assigns nothing, syncs with no project system and sends no status emails; it keeps what each stakeholder said, marked with where it came from.
Last updated 2026-10-28.
All product names, logos, and brands are property of their respective owners. Equerry is not affiliated with, endorsed by, or sponsored by any other product mentioned; comparisons are provided for informational purposes.