Google's Antigravity IDE shipped a complete productivity app landing page without leaving the editor, according to a Show HN post from developer ravi_rupareliya who connected two MCP servers inside the platform. GitHub MCP handled repo creation, branching, and pull request generation via natural language prompts. Stitch MCP pulled design tokens directly into generated code. The author ran the setup on a live project for several weeks — not a polished one-off.
The two integrations are at different stages. GitHub MCP is the more mature: repo management and PR generation with descriptions work smoothly through natural language alone. Stitch MCP is explicitly flagged as early-stage by ravi_rupareliya, which makes Google's design-to-code story still a work in progress.
What separates Antigravity from most <a href="/news/2026-03-14-modulus-parallel-ai-coding-agents">AI coding tools</a> is where MCP sits in the architecture. The majority of AI IDEs still function as enhanced autocomplete or in-editor chat. With MCP baked in at the platform level, Antigravity can chain tool actions — from commit to code review — without the user switching context or wiring up one-off plugins. Anthropic developed the MCP standard; it has since spread across editors and <a href="/news/2026-03-15-axe-go-binary-toml-llm-agents-unix-pipes">agents</a> as a shared integration layer. Antigravity's native support puts it in a short list of IDEs treating the protocol as core infrastructure rather than a bolt-on feature.
Stitch MCP's early-stage flag is the clearest indicator of where the workflow still has gaps. If Google closes that loop, the design-to-code step becomes the most compelling piece of the whole demo.