P
PulseTask Portal
Live in production on AWS

The portal a team opens first thing in the morning.

Tasks, hours, capacity, appraisals, quality gates and corrective actions - one login, one database, on your own server. I designed, built and deployed the whole thing myself.

A task can't reach Closed until its checklist is clear

In progress
TAN-T6Test the portal end to endChecklist 0/4
To be tested
TAN-T4Coordinate with fulfilmentOverdue
Closed
TAN-T3Track and update stock detailsChecklist 5/5
TAN-T5Finalise the seasonal sale offersChecklist 5/5

Operations checklist, version 4

  • Stock counted and reconciled
  • Approved by quality control
  • Rollback plan verified
  • Handover notes written

This week's booked hours

SunMonTueWedThuFriSat
8Modules, from the task board to appraisals
5Stages a task moves through before it closes
3Access levels: admin, manager, employee
1Server, one database, no per-seat licence

Why it exists

Three questions most teams can't answer without asking around.

Who is buried this week?

Hours lived in a spreadsheet nobody opened until payroll. Work went to whoever replied fastest, not whoever had room.

Is this actually finished?

"Done" meant different things to different people. Tasks closed with steps skipped, and the gaps surfaced weeks later as customer complaints.

Did we ever fix that?

Problems got discussed in chat, solved once, and forgotten. The same fault came back the next quarter with nobody able to say what was done last time.

What Pulse does

Eight parts of running a team, joined up.

Each one is useful alone. Together they mean a task closing updates someone's score, a checklist blocking a close raises a problem, and a week's logged hours redraws the capacity grid.

Task board

Five stages from Open to Closed, drag a card to move it. Filter by project or owner, switch to a list, see due dates turn red the day they slip.

Quality checklists

Every team has its own checklist and every task inherits it. The card will not move to Closed until each item is ticked. Checklists are versioned, so raising the standard never rewrites what was already signed off.

Timesheets

People log hours against the task they were actually doing. No separate tool, no end-of-month reconstruction from memory.

Capacity grid

A week at a glance: hours booked per person per day and a utilisation figure for the week. You can see who to hand the next job to before you promise a date.

Delivery score

On-time performance calculated automatically from closed tasks, per person and per team, with the full history behind every number.

KPI appraisals

Monthly targets with weightings, actuals logged with supporting files, and achievement, weighted score and rating worked out for you. Scorecards move through draft, submitted and approved.Editing a KPI that already has approved months files a new version that starts from the first open month. Signed-off appraisals stay exactly as they were.

Problems and CAPA

Root cause, corrective action and preventive action recorded against the task that caused them, with overdue corrective actions flagged so nothing quietly disappears.

People and teams

Roles, daily capacity, live status, and menu-by-menu permissions for each person. An employee sees their work; a manager sees the team.

How it was built

Built entirely with an AI agent running on my own machine.

Not a hosted assistant, not a browser tab. The coding agent ran locally, on my hardware, against a local copy of the project. That is why a system this size exists at all without a team behind it - and why nothing about it had to leave my network to get written.

Your code never leaves your side

Because the agent is local, source, database schema and any sample data stay on the machine. Nothing is uploaded to a third-party service to be stored, indexed or used for training.

Agency scope, solo speed

Eight modules with role-based access, a full appraisal engine and a production deployment - built and shipped by one person in a fraction of the usual time, and priced that way too.

Reviewed line by line

The agent writes fast; I read everything before it ships. Schema, access rules and anything touching money or appraisals get checked by hand. One person understands the whole system and answers for it.

local agent - pulse-portal

Inside the portal

Screens from the running system.

Not mockups. This is the portal in use, with names changed.

Capacity, a week at a time

Booked hours per person per day, utilisation for the week, filtered by team. Empty rows are as informative as full ones.

Pulse bandwidth screen showing a weekly grid of hours per employee with utilisation percentages

KPI scorecard with an approval trail

Targets and weights per role, actuals with attachments, achievement and weighted score calculated as you type, and a draft to approved workflow behind it.

Pulse KRI and KPI scorecard screen with targets, weights, actuals and weighted scores

Checklists that hold the gate

One checklist per team, each with its own version number and history. A task cannot close until every line is ticked.

Pulse checklists screen showing per-team checklists with version numbers

Problems, and what was done about them

Root cause, corrective action, preventive action, kept against the work that caused them. Overdue corrective actions are flagged, not buried.

Pulse problems and CAPA screen showing root cause, corrective and preventive actions

Who can see what

Role, daily capacity and status per person, with access granted menu by menu. Email addresses blurred here for privacy.

Pulse people screen listing employees with roles, capacity, status and per-menu access

Scores that come from the work itself

On-time delivery calculated from closed tasks rather than filled in by hand at review time.

Pulse scorecard screen showing total score, on time and late counts

What it runs on

Node 20 and Express with a React front end and MySQL, kept alive by PM2 behind Apache on an AWS EC2 instance - sharing a box with a live Magento store without either one going down. Email goes out over SMTP. You own the server, the database and the code.

Node 20 + ExpressReactMySQL 8PM2Apache reverse proxyAWS EC2 (Ubuntu)SMTP emailRole-based access

Your build

Pulse is the proof. Yours would be built the same way.

Most teams don't need all eight modules on day one. We start with the one screen that replaces your worst spreadsheet, put it in front of your people, and grow it from what they actually use.

Tell me the part of your week that runs on a spreadsheet and someone's memory. I'll tell you honestly whether software fixes it.