How we run Pulse as an open-source business — license design, governance, self-host economics.
Blog
Two a month.
Six months committed.
A publishing schedule, not a landing page promising "coming soon". Every
post below has a date, a track, a working title, and a concrete
takeaway. Subscribe by starring the repo — every post ships as a git
commit to apps/marketing/src/pages/blog/.
Tracks.
Financial ops for services firms: project profitability, utilisation, forecasting, close-the-month playbooks.
Building Pulse toward MCP + tool schemas so Claude, Copilot and other agents can ground on HR data safely.
TanStack Router, Bun workspaces, Astro marketing + React app monorepo, PWA offline, local intent parser.
Schedule.
-
#01
May 6, 2026
PlannedOSS mechanicsWhy we picked FSL-1.1-MIT (and not AGPL, BSL, or plain MIT)
A walk through four alternatives, the failure mode each prevents, and why a two-year converting license is the honest compromise for a funded open-source SaaS.
-
#02
May 20, 2026
PlannedCommessa opsThe commessa primer: why services firms need a project-first data model
Finance sees the company as a stack of commesse (billable project codes). HR tools model it as departments. This is the root cause of every "is this project profitable?" argument.
-
#03
Jun 3, 2026
PlannedEngineering notesMigrating off TanStack Start: a Vercel-first SPA in 24 hours
Why we left SSR behind for a vanilla Vite SPA, what broke, what we kept, and the exact vercel.json that ships today.
-
#04
Jun 17, 2026
PlannedAgent-native HRShipping a read-only MCP server in front of tenant data
What endpoints to expose first, how to scope permissions per tool, and the one security guarantee you cannot skip.
-
#05
Jul 1, 2026
PlannedOSS mechanicsThe free-for-5 economics: pricing an open-source HR SaaS
A spreadsheet you can fork. How the 5-employee free tier survives CAC, and what it costs us per free team in marginal infra.
-
#06
Jul 15, 2026
PlannedCommessa opsClose the month in an afternoon: the Saturation playbook
Step-by-step from timesheet lock to margin-closed-books, using Saturation insights to catch the three issues that blow most closes.
-
#07
Aug 5, 2026
PlannedEngineering notesA 700-line intent parser beats our first LLM call
The deterministic nlp.ts module behind ⌘J. Six intent kinds, fuzzy matching, confidence scoring, and the benchmarks that convinced us to ship it instead of calling Claude.
-
#08
Aug 19, 2026
PlannedAgent-native HRLetting agents approve expenses — a permission model that doesn't leak
Write-actions via MCP without broad delegation. Per-tool permission scopes, replayable dry-runs, and an audit surface that makes compliance happy.
-
#09
Sep 2, 2026
PlannedOSS mechanicsSix months of public issues: what the community actually wants
Every GitHub issue opened in the first six months, clustered by theme, with the three features we'll ship because of them and the two we said no to.
-
#10
Sep 16, 2026
PlannedCommessa opsScenario planning on your own burn: the Forecast walkthrough
"What if I add a designer in August?" Scenario sliders walked through with numbers and the kinds of decisions they actually change.
-
#11
Oct 7, 2026
PlannedEngineering notesSelf-hosting Pulse on a €25 Hetzner box
Docker Compose file, step-by-step, runbook for a 50-person team. Benchmarks, backup rota, and the two things you must not skip.
-
#12
Oct 21, 2026
PlannedAgent-native HRFrom command bar to agent loop: what changes when the user isn't human
The UX and product implications of a tool surface agents call directly. What breaks, what becomes possible, and the new security failure modes to design against.
Want to write for us?
Guest posts welcome — especially from HR ops leads, fractional finance folks running commessa-heavy consultancies, and engineers running their own self-hosted deployment. We pay €400 per published piece (€200 for the draft even if we don't run it) and every guest byline links back to whatever you're building.
Pitch an idea → editorial@pulsehr.it
Your team deserves
better software.
Free for the first 5 employees — forever. No credit card. Import your data in under an hour.