Taskosaur

Taskosaur's core claim is narrow and testable: AI Mode executes. The homepage walks through the distinction carefully — traditional AI features answer "You might want to create a task," while Taskosaur's AI Mode reports back with checkmarks against real state changes: found 12 P0 bugs, created "Bug Fix Sprint #3," assigned by expertise and workload, team notified. That execution loop, plus event-triggered workflow automation ("when design tasks finish, auto-create dev tasks"), is where the product earns its keep. The rest is familiar PM plumbing — boards, sprints, role-based permissions, smart filtering, resource balancing, template-driven workspace setup — wrapped in an interface where the fastest path is a sentence. The economics are the second pillar. Self-hosting means no per-seat fees, and bringing your own AI keys means you pay your model provider directly instead of a vendor markup; the vendor's own example puts a typical team at $12K/year against $120K/year for seat-priced SaaS PM. If you already run infrastructure, that gap is real money, and the data-control argument — three years of project history on your own servers rather than trapped in someone's cloud — is the kind of thing that matters most precisely when you want to leave, which is when SaaS vendors are least helpful. What you are accepting in return: Taskosaur is early. Deployment is on you (Docker or your own infra), and it assumes DevOps comfort rather than offering a managed path in the material provided. AI reliability tracks your chosen model's rate limits and latency because those are your keys. There are no native mobile or desktop clients, so anyone who works from a phone is out of the conversational loop. Compliance is the sharpest edge: SOC2 and GDPR are described in the seed material as "ready" rather than certified, which for a regulated buyer is the difference between a checkbox and a blocker. Finally, BSL is source-available rather than a permissive OSI license, so if your legal team treats "open source" strictly, read the license before you plan a fork. Where it fits: engineering-led teams of roughly 5 to 50 who already self-host, agencies juggling multiple client workspaces that benefit from instant template setup, and privacy-conscious groups that cannot put project history in a third-party cloud. Where it does not: non-technical teams, mobile-first organizations, and enterprises that need certifications on day one. The honest framing is that Taskosaur trades vendor risk for operational risk — you are betting on your own ops competence instead of someone else's SLA, and for the right team that bet pays out in both dollars and control.