For developers and small agencies building client projects with AI.
Finish 8× more tasks with Projectous than using Claude Code alone.
Remove the AI bottleneck. Your ideas, development, and reviews can move forward at the same time. Projectous keeps every session focused on the right work.
14 days. No card.
You stop writing tasks
Claude is already holding the context the work needs, so it writes the task and you do not. One line while you are talking becomes a brief complete enough that a session which was not there can pick it up and finish it.
- 01
You say it in a sentence
“file a task for the badge that does not clear when you read a chat somewhere else, and give it to Richard”. That is the whole input. No form, no second window, and no stopping the thing you were in the middle of.
- 02
Claude writes the task, you do not
It knows the file, the error and why the obvious fix is wrong, because it was just there. All of it goes into the task. You would have had to type that out yourself, and now you do not.
- 01
A session takes the next one on its own
By priority, rather than by whichever one you happened to remember. Taking it holds it, so a second session cannot start the same work.
- 02
It starts cold and still finishes
The brief carries the context, so a session that was not there when you had the idea does not need you to explain it again. This is the whole reason intake matters. A backlog of one-line reminders is not something anything can pick up.
- 03
Checks decide when it is done
Gates you define in your own repository, not in somebody else's head. The agent opens the pull request and marks the task when it merges.
- 01
Merging leaves you something to check
The gates write what to verify onto the task itself, so a published change arrives with its own test rather than as a line in a changelog. A badge in the app counts how many are waiting.
- 02
You review
Open the ones that are waiting, confirm them in production, and either mark them done or send them back. Reviewing is the job. The bookkeeping around it is not.
# once, when you set the repo up
$ npx @projectous/pt init
ok credentials written
ok pointer added to your agent's instruction files
# every day after that, you just talk
you the drawer badge does not clear when you read a chat
somewhere else. and while you are in there, the project
chat opens the wrong panel
agent filed both. taking the badge one first because it is
the one people notice.
agent done. 16 gates passed, pull request open, task stamped
with what to check once it deploys.
# later, in the app: 1 waiting to be checked
# you open it, try it, and mark it done or send it backSetup is one command per repo and one sign-in per machine, and the package is published. [MISSING. Check the agent’s side of that exchange against a real session transcript, so the wording matches what an agent actually says.]
Evidence
More work finished, by fewer people, in less time each
One workspace, the one this product is built in. In one week of June 2026 three things landed: the console the agent drives, the gates that decide when work counts as done, and the habit of writing every trap into the repository instead of re-explaining it. Everything below compares the 171 days before that week to the 113 days after.
One more number, because it is the one a developer asks for. Reverts ran at 11 in 1,059 commits before that week and 14 in 2,788 after, so the revert rate halved while commit volume nearly tripled. Fixes per feature fell from 2.0 to 1.6 over the same split. Both point the same way. Neither isolates the gates from the console or from the written-down traps, because all three arrived together.
What it is not
- Not AI arriving. Claude had been in daily use for more than a year by then, and the rate sat at five a day throughout.
- Not the published package. That was 1 July, eleven days after the step.
- Not a busier month. The first nineteen days of June ran at 4.9 a day. The last eleven ran at 32.5.
- Not one change. The console and the gates arrived on the same day in adjacent pull requests, and the first written-down trap the day after, so nothing here can tell you which of the three did the work.
The complaint about AI is usually the review queue: it produces more than a team can get through, so somebody ends up digging. The opposite happened here. The same people finished three times as much, and a higher share of it inside a month.
Queried from production and from git history on 11 October 2026Companies using Projectous
The work arrives faster than the memory of it
“If I didn’t have Projectous I wouldn’t be able to track all my ideas in Claude sessions. Literally thousands of tasks over the last few months have been created and processed.”
Stéphane Deuvaert, who builds Projectous in Projectous
- A session ends with four good ideas in the scrollback and none of them anywhere you will look again.
- You fix the bug you remember rather than the one that matters, because the list of both lives in your head.
- A client asks where something stands and you go reading chat history to find out.
- The thing you promised on Tuesday is still true on Friday and nothing anywhere knows it.
Across every repo you run
A lesson learned on one project turns up on the next one
Checks start local. When one of them turns out to be true of any codebase rather than just yours, it moves up into the shared set, and every repo you have picks it up on its next run.
- You explain a trap once and your agent writes it into the repository, so the laptop you pick up tomorrow and the person who clones it next week start from everything you have already said. In this one that is 59 traps and 46 architecture notes, all committed.
- A client project teaches the shared set something, and the next project you start already knows it.
- Anything genuinely specific to one codebase stays in that codebase. The rule is local by default, shared only when the lesson holds anywhere.
- A fresh clone catches itself up. If the copy you pulled is behind the shared set, the first run brings it current rather than running the old checks.
- Reading-heavy checks run as separate agents that hand back a short summary, so the session you are talking in does not fill up with files it only needed to glance at.
[MISSING. Model tiers are configurable per project and are not switched on in the workspace these numbers come from, so nothing here claims a cost saving. The shared set is also ours, so decide what this section may promise a reader before it takes traffic.]
Not a single player tool
Solo today does not mean solo in March
You are already two: you and the agent. The system is built for that and keeps working when the third and fourth arrive, whether they are people, their agents, or the client.
- Two conversations on the same task. One your side can see and one the client can, so you can argue about the estimate in the first and answer the client in the second without leaving the work.
- Nobody starts the same task twice. Taking a task holds it, and a second session asking for the same one is told it is taken rather than quietly duplicating a day.
- Everyone signs in as themselves, so a change is attributed to the person whose agent made it rather than to whoever set the repo up.
- A new person does not need you to re-explain the project. What you wrote down once is in the repository they just cloned, and their agent reads it on its first session.
What it is like once the work has somewhere to live
“Projectous has everything in one place. You can sort and search tasks fairly easy. I can add images and files straight away on every task or project.”
Jabril King-MahdiCSM, Rhymes with Reason“Projectous has been the best system that we have been using from a developer's perspective. Trello and Asana got messy real quick.”
Austin MartinFounder & CEO, Rhymes with Reason“Hyperfind was struggling to keep the workflow streamlined using multiple tools like Trello, Asana, Hubspot, QuickBooks. We were disorganized to the brink of failure. Projectous basically saved us by streamlining the whole agency process — everything under one roof.”
Durim B.Co-Founder & CEO, Hyperfind[MISSING. None of these quotes carries a number. Ask two of them what changed in hours or tools dropped.]
Who this is not for
- Your first app, built on weekends. The problem here starts when you have more ideas than hours, and if you do not yet, this is weight you would carry for nothing.
- A team that wants planning depth. No Gantt chart and no resource planning, on purpose.
- Anybody who wants their tasks to live inside one chat window. These outlive the session deliberately, which is the whole point and also the cost.
Price
Every number, on the page
Solo
- 1 user, and $7 a month for each one after
- The console, the gates and the Today list
- Projects, tasks, timers and the invoice at the end of them
- A password vault, encrypted on your machine before it is sent
Cancel whenever. [MISSING. Confirm the cancellation wording against what billing actually does.]
The five you are about to ask
Does signup make me set up an agency before I can do anything?
[MISSING. This is the one that decides whether the page works. If a solo builder lands here and signup asks them to invite a team and configure clients, the promise breaks on the first screen. Answer it honestly, or fix signup before running traffic.]
Does it work with Claude Code specifically?
Setup writes a pointer into the instruction files Claude Code reads at the start of a session, so it picks the workflow up without being told again. Anything that reads those files and can run a command works the same way. Projectous itself is built like this.
What happens to my tasks if I stop paying?
[MISSING. Export and retention are a real objection for somebody about to put thousands of items in. Say what actually happens.]
How long before it is useful?
[MISSING. Measure one real first session rather than guessing. The claim only needs to be true, not short.]
I have clients too. Is that in the way?
The client side is there when you need it, with its own chat, contracts and invoices built from tracked time. Nothing asks you to use it to keep a task list.
Put the next idea somewhere it will still be tomorrow
Start my trialHe built it for this exact problem.
Talk to Stéphane instead