
The MCP connector changed everything for us. We're a nonprofit (Quietly Working Foundation) running 9 apps, 140+ automation scripts, and 7 core operational workflows... and we needed a way to make all of that visible. Not just to me... but to students learning how operations actually work.
Puzzle gave us that. We connected the MCP to our AI agent, and in one session built 7 complete workflows with 77 steps, role assignments, and tool references... programmatically. No dragging boxes around for hours. The agent reads our SOPs and writes them into Puzzle as living process maps. That's the AI integration working exactly how it should... not a gimmick, but a real accelerator.
The canvas UI is genuinely beautiful. Clean layout, intuitive grouping with sections, and the way tools and roles attach to steps makes complex operations feel approachable. Our students can see how a YouTube video becomes a published article across 9 steps, who's responsible for each one, and which tools power it. That's not a diagram... that's a teaching tool.
Support has been outstanding. We reported a bug with the create_process MCP endpoint and Jolly from their team escalated it to engineering... fixed within 24 hours. Brian Ragone (founder) personally followed up. That kind of responsiveness from a product team is rare.
Performance is solid... the MCP handled 7 workflow creates with 77 steps and 66 connections without breaking a sweat. Canvas loads fast even with 20+ sections visible.
Unlimited workspaces means we can eventually give every supporter their own operational blueprint. For a nonprofit trying to scale operations training... that's the dream. Review collected by and hosted on G2.com.
Honestly... not much. But if I'm being helpful...
The MCP connector uses OAuth that requires re-authentication every chat session. For teams running AI agents on remote servers, a static API token option would make the integration much smoother. I've submitted this as a feature request and their team acknowledged it quickly.
Tab management isn't available through the MCP yet. You can filter by tab when reading data, but you can't create tabs or move sections between them programmatically. When you're building 7 workflows via API, they all land on one canvas tab and you have to organize them manually in the UI afterward. Not a dealbreaker... just a gap between what the MCP can build and what the UI can organize.
Dark mode isn't available yet... though their team told me it's on the public roadmap. For someone staring at process maps for hours, that matters.
That's really it. The product does what it promises and the team is actively building. These feel like "not yet" items, not fundamental problems. Review collected by and hosted on G2.com.