Open-Source SaaS-Killer Stack, Part 3
Same eleven tools, two different companies. MySeoDesk is my own company and already self-hosts most of this stack on Contabo. J.M. Field is my employer, with its own corporate infra and none of this running yet, so every per-seat saving is still on the table.
Sources: "10 GitHub Repos So Good They Shouldn't Be Free, Part 3" (Hyperautomation Labs) and the NotebookLM agentic update (WorldofAI). The video claims roughly $180k a year of replaced software across all thirty part 1, 2, and 3 repos. Free PDF of all 30 with deploy commands.
Context. MySeoDesk is a solo-operator agency self-hosting on Contabo. Already running: Vaultwarden, DocuSeal, Mailcow, NotebookLM, Obsidian, Figma MCP. So the play is to deploy the gaps, layer onto existing infra, and skip anything that duplicates what I own.
Biggest upside. Drag-drop internal tools bound to my Postgres and FastAPI APIs: client ops dashboards, audit-job trackers, deploy consoles. Customer data never leaves Contabo.
Fills a real gap, since I run no CRM today. Notion-style pipeline for client and prospect deals, custom objects for engagement types, webhooks into my sites. Catch: AGPL, and some enterprise features are missing.
Layer it on top of my existing Mailcow or SES as the SMTP backend, a campaign layer on infra I already own. SQL segmentation, live opens, clicks, bounces. Honors the one-rollup, no per-alert-blast rule. I already manage deliverability.
Client walkthroughs, MySeoDesk audit and demo videos, async updates. Recordings stay on my servers. The Mac and Windows desktop app works out of the box; self-host the sharing stack with Docker or Railway.
Complementary, not a replacement. Looker Studio stays my client-facing GA4 and GSC reporting; Superset becomes internal ops BI on my own database. Catch: heaviest setup of the ten.
Enabling, not cost-cutting, since I run no managed vector DB today. Reach for it when I build semantic search over client catalogs, or memory for the agent stack.
Use the new agentic update for client-brief synthesis and competitor research, exported as decks (PDF, DOCX, XLSX, PPTX) with source attribution. Keep going.
Already self-hosted on the mail VPS. No action, just route every contract, SOW, and client sign-off through it.
Already self-hosted. No action. Backups and lockdown are already mine to own.
Do not rule this out. Figma stays primary for one concrete reason: I just wired the Figma MCP into client builds for design-to-code, and Penpot has no equal to that yet. But Penpot is the only option that fits my self-host ethos. Files live on Contabo next to Vaultwarden and Mailcow, seats are unlimited and free, and the handoff exports real CSS because it is built on open web standards. It is the hedge against Figma price hikes and lock-in, and the cost to try is one Docker command. Trade-off: smaller plugin ecosystem, and a few power features such as deep variants and advanced auto-layout still lag Figma.
Redundant. My Obsidian vault is already the second brain. Migrating a working system buys nothing.
Context. J.M. Field is my employer: fulfillment, commercial print, and marketing, B2B, with a real team. I do not own the infra. Their corporate stack runs on Enhance or cPanel and on-prem Exchange or M365, and none of these tools run yet, so the full per-seat savings apply. Treat this list as recommendations to bring to JMF IT. Adoption needs their buy-in, not a unilateral deploy.
B2B pipeline for fulfillment, print, and marketing accounts, with custom objects for print-order types and API or webhook ties into JMF sites. Catch: AGPL, and some enterprise depth is missing.
The giant-killer for an ops shop. Dashboards on print turnaround, cost per order, fulfillment SLAs, and kitting throughput, straight off the warehouse database. Company-wide BI for the price of a server.
Client contracts, POs, NDAs, and vendor onboarding. Drag fields onto a PDF, audit trail, tamper-proof completed document. Easiest deploy in the set, a coffee break. Catch: leaner integrations than DocuSign enterprise.
Internal fulfillment and order-status tools, kitting and inventory consoles, support panels, all bound to JMF databases. Customer data stays on JMF servers, the real reason ops teams switch.
JMF has an actual creative and marketing design team. Unlimited self-hosted seats, dev-handoff to clean CSS, files kept in-house, which regulated client work likes. Catch: smaller plugin ecosystem.
Marketing-arm newsletters and client campaigns at list scale. SQL segmentation, live opens, clicks, bounces, plugged into SES. Catch: JMF IT owns sending reputation, which Mailchimp babysits.
Client demos, warehouse SOP and training videos, async stand-ups. Self-host so customer recordings stay off third-party servers.
Internal docs and wiki plus a whiteboard for campaign planning, on one local-first canvas. Catch: real-time multiplayer is still maturing versus Notion.
Team password vault, unlimited users, orgs, and 2FA. Official Bitwarden clients work against it. Catch: JMF IT owns backups and hardening, which must be done properly.
Only if JMF builds AI search or retrieval: product-catalog search, a knowledge base. No saving until there is a use case.
Research and proposal generation for the marketing team: synthesize briefs into client-ready decks with source attribution. A subscription cost, not a savings play.