Christmas Day: 7 Open-Source Tools the Softechinfra Team Used Most in 2025

Christmas Day in Delhi is quiet. The team is mostly off, the standups are on hold, and the only things still running are the seven open-source tools we leaned on hardest in 2025. This is the Softechinfra team's honest favourites list — what we used, what it replaced, what it cost (mostly nothing), and one line on why we kept coming back to each. Order is rough usage frequency across our internal stack, not a ranking of software quality. ## TL;DR — the seven tools in one paragraph n8n for workflow automation. Plausible Analytics for cookie-less, GDPR/DPDP-friendly site analytics. Excalidraw for sketches in client calls. Cal.com for booking and scheduling. Uptime Kuma for status-page monitoring. Posthog for product analytics on internal tools. NocoDB as our airtable-replacement spreadsheet/database hybrid. All seven self-hosted on five small Hetzner CX22 instances. Total infra: under ₹4,000/month for tools that would have cost ₹20,000+ on equivalent SaaS plans. ## Why this matters now (and why a holiday post is the right time) Christmas is when teams pause and re-evaluate the stack. December is also when annual SaaS renewals start landing in inbox. The honest case for open-source, self-hosted tools in 2025 is not ideological — it is operational. The big four reasons we keep choosing them: data lives on our infra (DPDP Rules notification this November made this a board-level concern), unit costs at scale are 5-15x cheaper, customisation is genuinely possible, and the community templates have hit critical mass for tools like n8n. The big tradeoff: ops time. Roughly one engineer-hour per week to keep five Hetzner instances healthy. Worth it. ## 1. n8n — the workflow automation we run on every project What we use it for: 14 internal workflows including new-lead Slack notifications, weekly client-report generation, Calendly-to-CRM-sync, GST-invoice email parsing, Hetzner backup health-check pings, and an LLM-summary pipeline that condenses our weekly engineering Slack into a Monday digest. Plus client-side: every AI automation engagement we run for clients ships on n8n. What it replaced: Make.com (which we still use for one-off short-lived workflows where speed-of-setup beats cost) and Zapier (we sunsetted in March 2025). Why we use it: n8n self-hosted is free with no licence cap on executions. The community has hit critical mass — there are 280+ ready templates on the awesome-n8n-templates repo and over 9,652 templates on n8n.io's community page. AI nodes (LangChain integration, OpenAI, Anthropic) shipped first-class in 2024 and matured through 2025. Real cost: one Hetzner CX22 at ₹740/month runs all 14 workflows comfortably. ## 2. Plausible Analytics — cookieless, fast, DPDP-compatible What we use it for: Site analytics for softechinfra.com plus three client sites where the marketing team wants live dashboards. What it replaced: Google Analytics 4 (which we still use for two clients who specifically need GA's audience features, but the default for new builds is now Plausible). Why we use it: No cookies, no consent banner, GDPR/DPDP-compatible out of the box. Page-load impact is ~1.5KB vs GA's ~46KB. Real-time dashboard with the metrics we actually look at — pageviews, sources, devices, top pages — without the 200-metric cognitive load of GA. With DPDP Rules now notified, the consent-banner-free workflow is a real operational win. Self-host on the same Hetzner box that runs n8n. ## 3. Excalidraw — every team call ends with one of these What we use it for: System architecture sketches in client kick-off calls, design wireframes when Khushi needs a quick low-fi mockup, and the whiteboard moments in our weekly engineering review. What it replaced: Miro (still in use for two large-team clients but Excalidraw covers 80% of internal need), and physical whiteboard photos. Why we use it: The hand-drawn aesthetic communicates "this is a sketch, not a spec" — which is exactly the right vibe for first-call architecture conversations. Free, zero-config, exportable to PNG/SVG/Excalidraw native. Works offline. The integration with VS Code (excalidraw extension) means architecture diagrams live next to code. We export to SVG and embed in blog posts when explaining system flow. ## 4. Cal.com — booking that does not embarrass us in front of clients What we use it for: Discovery-call booking on the /contact page, internal team availability sharing, and the founder-call calendar that Vivek uses for first calls. What it replaced: Calendly (we still pay for one Calendly seat where a specific Calendly integration matters; everything else moved to self-hosted Cal.com in mid-2025). Why we use it: Same UX as Calendly, fraction of the cost, plus full control over branding. The OSS version handles round-robin, team event types, payment integration (Stripe), and webhook triggers into n8n. Routing forms (the booker answers a few questions before seeing the calendar) is the killer feature for filtering low-fit leads before they land on the founder's calendar. ## 5. Uptime Kuma — the cheapest status page on the planet What we use it for: Monitoring our own client-facing dashboards (16 endpoints across 9 client projects), plus internal services. Pings every 60 seconds, alerts to Slack and email when anything misses. What it replaced: A combination of Pingdom (we cancelled the Standard plan in April 2025), Better Uptime (we never adopted), and the manual "did anyone notice the staging server is down" Slack messages. Why we use it: Self-hosted, beautiful UI, public status page for clients who want one, supports HTTP/HTTPS/TCP/Ping/DNS/Docker/keyword monitoring. About 20 minutes to deploy. Real cost: zero — runs on the same Hetzner box as n8n. The Slack alert routing through our incident channel means the on-call engineer knows about an outage in seconds, not minutes. ## 6. PostHog — product analytics on internal tools What we use it for: Behavioural analytics on internal tools we built for ourselves (the project-tracker, the time-log, the AI-prompt library) plus product analytics for two client products we maintain. Funnel analysis, session replay, feature flags, A/B tests. What it replaced: Mixpanel (still used by two enterprise clients we maintain) and a custom-built event log we sunset in 2024. Why we use it: Self-host or use the EU-hosted free tier (1M events/month, generous). Session replay is genuinely good — we caught three major UX bugs in our internal AI-prompt library in 2025 that we would never have found from heatmaps alone. Feature flags integrated with the same product means experiments are one config change, not a new deploy. ## 7. NocoDB — the Airtable replacement for internal data What we use it for: Project tracker, client-CRM-lite for the discovery-call pipeline, content calendar for blog posts, and the leave-tracker the operations team maintains. Five distinct internal databases. What it replaced: Airtable (we sunset our paid Airtable workspace in February 2025). Why we use it: Looks and feels like Airtable, runs on Postgres (so the data is portable and SQL-queryable), free for self-host. Forms, Kanban, Gallery, Calendar views all present. API and webhook support means it integrates with our n8n workflows cleanly. The migration from Airtable was about 3 hours per base. Total saved: ~₹14,000/month in Airtable fees that now goes nowhere. ## The cost summary ## The honest tradeoffs We do not pretend this is free. Three real costs. Ops time: Roughly 4-6 engineer-hours per month across all five Hetzner instances — patching, backup health-checks, occasional minor incident. At our internal hourly rate that is ~₹4,000/month. Net saving versus SaaS is still strongly positive but not "zero." Onboarding cost: New team members need a Hetzner SSH key, the runbook for "where is X tool," and a half-day pairing session. SaaS tools are click-button onboarding. We absorb this as part of standard new-hire onboarding. Failure-mode complexity: When a Hetzner instance restarts, three tools may go down briefly. Uptime Kuma alerts us. The handful of incidents we have had in 2025 averaged ~12 minutes mean-time-to-recovery. Acceptable for internal tools, would not be acceptable for client-facing services. ## What we tried and dropped this year For honesty: three open-source tools we adopted briefly in 2025 and dropped. Mautic (open-source marketing automation). Powerful but heavy. The setup-to-value ratio did not work for our scale — we ship marketing automation for clients on n8n now and only consider Mautic for clients with 100k+ contact lists. Outline (open-source wiki). Good product. Lost to Notion which the wider team had stronger habits with. We may revisit if Notion ever ships a pricing change we cannot stomach. Plane (open-source project management, Jira alternative). Felt promising but our Linear and ClickUp habits across the team were too entrenched to switch. Recommended to two clients who were starting fresh; both are happy. ## Our 2025 stack at a glance

Tool	Use case	What it replaced	Monthly saving (est.)
n8n	Workflow automation	Make + Zapier	₹4,800
Plausible	Site analytics	GA4 partially	~₹0 (GA is free)
Excalidraw	Sketches	Miro partially	₹2,400
Cal.com	Booking	Calendly	₹1,800
Uptime Kuma	Monitoring	Pingdom	₹2,800
PostHog	Product analytics	Mixpanel partial	₹5,700
NocoDB	Internal databases	Airtable	₹14,000

• n8n live on Hetzner — 14 internal workflows, all client AI-automation builds
• Plausible deployed for softechinfra.com + 3 client sites
• Cal.com handling all internal + founder-facing booking
• Uptime Kuma monitoring 16 endpoints across 9 client projects
• PostHog instrumenting 5 internal tools + 2 client products
• NocoDB hosting 5 internal databases, Airtable sunset
• 5 Hetzner CX22 instances at ₹740/mo each = ₹3,700/mo total infra

## A real example — what we built TalkDrill on this year TalkDrill, our in-house English-fluency app, runs in production on a stack that is roughly 60% open-source. PostHog handles funnel analysis on the onboarding flow. n8n powers the daily-summary email pipeline. Plausible runs the marketing site analytics. The stack scaled from 500 to 5,000 active users over 2025 without touching the open-source choices. The cost story is real: TalkDrill's monthly tooling spend stayed roughly flat through 10x user growth, while the equivalent SaaS bundle would have grown about 4x. TalkDrill is one of two in-house products (the other is PenLeap) where we stress-test our internal stack at real-user scale before recommending it to clients. ## Common mistakes when adopting open-source self-hosted Underestimating backups. Self-hosted means you own the backups. Set up automated nightly snapshots from day one — Hetzner snapshots cost ₹50-₹100/month per server and have saved us twice in 2025. Skipping the upgrade routine. Open-source projects move fast. Schedule a monthly "patch Friday" where one engineer runs the upgrades. Skipping months of upgrades makes the eventual jump painful. Going all-in too fast. Migrate one tool at a time. Run the SaaS and the self-hosted in parallel for 2-4 weeks. Sunset the SaaS only when you are sure. Ignoring the bus factor. If only one engineer knows how to ssh into the Hetzner box, you have a single point of failure. Document the runbook, share SSH access with at least two people. Treating open-source like SaaS. No SLA. No 24x7 vendor support. No guaranteed roadmap. If your business cannot tolerate any of those, stay on SaaS for that specific tool. ## FAQ — what teams ask us when they see this stack ### Should I self-host all of these or pick a few? Pick three to start. n8n for automation, Cal.com for booking, Uptime Kuma for monitoring is a low-risk starter trio. Add the others in quarter 2 once your ops habit is built. ### Where do you host? Hetzner Cloud (Germany or Finland). Five CX22 instances at ₹740/mo each. We have evaluated DigitalOcean, Vultr and Linode — Hetzner is consistently the best price-performance for European-hosted workloads. For India-resident-data workloads we use AWS Mumbai with the same pattern. ### How do you manage backups? Hetzner snapshots scheduled nightly via the Hetzner API + a weekly off-site sync to AWS S3 Glacier. Total backup cost: ~₹400/mo across all five servers. ### What about security and patching? Unattended-upgrades for OS-level patches. Manual review of major-version bumps for each tool monthly. We use Tailscale for SSH access (no public SSH ports open). UFW + fail2ban as baseline. ### What is your Slack alert workflow? Uptime Kuma → webhook → n8n → Slack channel #alerts-prod with @here mention if HTTP-status is 5xx. PagerDuty-style escalation only for our two client-facing critical services. ### How long did the full migration take? About 8 months from "let's try n8n" (April 2024) to "we have sunset 70% of paid SaaS" (December 2024). 2025 was the consolidation year. ### Where do I find ready templates? For n8n, the awesome-n8n-templates GitHub repo (280+ workflows) and the official n8n.io workflows page (9,652+). For Cal.com, the official Cal.com app store. For everything else, the GitHub README is your friend.