Caret vs Lindy AI MCP Server Marketplace

“Hooking into the macOS Accessibility layer for universal autocomplete is exactly the right architecture — no app-specific plugins, no context-switching. If the latency is under 200ms this is an instant productivity multiplier for anyone who types for a living.”

“The primitive here is a hosted MCP client that resolves server discovery and auth so you don't have to — that's legitimately useful friction removal. But the DX bet is that no-code is the right layer for agent integrations, and that's exactly where I get off. MCP is a protocol designed so developers can compose tools programmatically; putting a marketplace UI on top of it doesn't make agents more capable, it makes the configuration surface bigger and the debuggability worse. The moment-of-truth test: when your agent misbehaves at step 4 of a 6-step workflow, how do you trace which MCP server returned bad data? If the answer is 'check our logs dashboard,' I'm reaching for the raw SDK every time.”

“Accessibility API access is a significant permission to grant any app — this tool can see everything you type in every application. Until there's a clear privacy audit and local model option, the security surface is hard to accept for professional use.”

“The category is no-code agent integration, and the direct competitors are Zapier's AI actions, Make's AI modules, and n8n's MCP nodes — all of which have larger connector libraries, more mature error handling, and existing user bases who already paid for the platform. Lindy's specific bet is that MCP standardization collapses the integration layer enough that being early to a marketplace wins, but MCP adoption among enterprise SaaS vendors is still thin enough that '150 servers' likely means 100 wrappers around the same REST APIs everyone already has. What kills this in 12 months: Anthropic ships native MCP tooling inside Claude.ai for Teams, and Lindy's marketplace becomes a curiosity for the 40 people who were using it.”

“System-level AI input layers are the next frontier after app-level AI. Caret is the first credible Mac implementation — expect Apple to build this natively into macOS within 18 months, validating the concept while commoditizing this specific product.”

“The thesis is falsifiable: by 2027, MCP becomes the TCP/IP of agent-to-tool communication, and whoever controls discovery and credentialing for that layer controls enterprise agent adoption. The dependency that has to hold is that MCP doesn't fragment into vendor-specific dialects the way REST+OAuth did — and that's a genuine risk, not a vibe. The second-order effect that nobody is talking about: if MCP server marketplaces win, SaaS vendors stop building native AI features and start publishing MCP servers instead, which quietly shifts the AI integration budget from the SaaS vendor to the orchestration layer. Lindy is early on this trend line — MCP standardization is six months old — and being early here means the catalog quality is thin, but the positional bet is real infrastructure thinking, not trend-chasing.”

“As someone who writes across Notion, Figma, email, and Slack simultaneously, a context-aware Tab that works everywhere is the dream. No mode-switching, no copy-paste to an AI chat window — just inline continuation of your own voice.”

“The buyer is a mid-market ops or RevOps lead who wants automations without an engineering ticket — that's a real budget and a real buyer, but Zapier already owns that person's credit card and their trust. Lindy's moat argument would have to be 'MCP-native from the start gives us better agent quality than bolted-on competitors,' but that's a technical claim dressed as a business moat, and technical leads evaporate when the better-funded player catches up. The pricing structure also doesn't scale with value delivered — flat monthly tiers for agent workflows mean your heaviest users are your worst unit economics, and 'contact sales' for business plans from a product this early signals they haven't figured out what enterprise customers actually need from this yet.”