All servers / AI / Antigravity MCP server
Antigravity MCP server
Google Antigravity MCP Server connecting any MCP client to Google Antigravity (Gemini)
Tested by ConnectATDScore 86 out of 100. Tested 2026-11-19 on package version 1.0.7. Package: @h0wzy/mcp-server-antigravity · source
Score breakdown
| Runs | 18 / 20 | |
|---|---|---|
| Tool descriptions | 20 / 20 | |
| Package safety | 20 / 20 | |
| Maintenance | 20 / 20 | |
| Adoption | 8 / 20 |
What we saw
- Starts and answers the MCP handshake: yes (1.6 seconds including download)
- Tools offered: 6
- Last published: 2026-10-08
- Weekly downloads (npm): 611
- License: MIT · Direct dependencies: 1
Want this server set up for your team?
ConnectATD builds AI automation and wires up MCP servers like this one. Tell us what you are trying to do.
Talk to ConnectATDTools and how they are described
Agents choose tools from these descriptions. Under each name you can see how many of its parameters are explained.
configure_antigravity4 of 4 parameters describedInspect or change Google Antigravity model and reasoning effort for the current session. Supports pre-configured tiers ("light" for trivial tasks, "balanced" for routine work, "deep" for hard bugs/complex architecture/security), explicit models (e.g. "gemini-3.8-flash", "gemini-3.1-pro"), or custom efforts ("low"|"medium"|"high"|"xhigh"|"max"). Actions: "get" (view active settings), "set" (apply updates), "reset" (restore startup defaults), "list" (catalog & tiers). Claude Code may switch tiers autonomously based on task difficulty.
ask_antigravity4 of 4 parameters describedGet an INDEPENDENT second opinion or answer from Google Antigravity (default model: Gemini 3.8 Flash High; configure via configure_antigravity). A different model family than Claude or OpenAI, ensuring an unbiased cross-check. Provide a `prompt`; optionally pass `paths`, `model`, or `effort`. Note: Antigravity also loads every MCP server the user configured in agy and can call their tools without asking (agy has no per-call switch), so only use it where those servers are safe to reach.
review_antigravity4 of 4 parameters describedRequest a thorough, structured code review from Google Antigravity (default model: Gemini 3.8 Flash High). Inspects code correctness, edge cases, race conditions, security vulnerabilities, performance, and architecture. Use tier "deep" via configure_antigravity for complex security/architecture reviews. Read-only: it cannot edit files, and shell commands the user has not allowed are skipped, so pass the files to review in `paths`. Note: Antigravity also loads every MCP server the user configured in agy and can call their tools without asking (agy has no per-call switch), so only use it where those servers are safe to reach.
brainstorm_antigravity4 of 4 parameters describedArchitectural brainstorming and exploration with Google Antigravity. Explores alternative design patterns, trade-offs, scalability considerations, and pros/cons. Configure model and effort via configure_antigravity. Note: Antigravity also loads every MCP server the user configured in agy and can call their tools without asking (agy has no per-call switch), so only use it where those servers are safe to reach.
plan_antigravity4 of 4 parameters describedGenerate a step-by-step implementation plan or execution checklist using Google Antigravity. Configure model and effort via configure_antigravity. Note: Antigravity also loads every MCP server the user configured in agy and can call their tools without asking (agy has no per-call switch), so only use it where those servers are safe to reach.
delegate_antigravity6 of 6 parameters describedHand a self-contained implementation task to Google Antigravity, which EDITS FILES inside `cwd` (permissions skipped). Review the diff afterwards. Returns the final report from Antigravity. Configure model and effort via configure_antigravity. Note: Antigravity also loads every MCP server the user configured in agy and can call their tools without asking (agy has no per-call switch), so only use it where those servers are safe to reach.
Tested in a sandbox with an empty environment: no API keys, a throwaway folder, nothing from a real account. A passing test shows the server runs and responds. It is not a security audit.