Loading ToolBench
MCP bridge that lets Claude Code delegate heavy tasks to the Antigravity CLI (agy) — purpose-built tools, model routing with fallback, and multi-turn session continuity.
agy-bridge is a delegation tool that outsources heavy tasks to the Antigravity CLI. All 6 tools are explicitly defined with Zod schemas, descriptions, and prompt builders. Naming is mostly clear and action-focused (analyze_*, deep_search, web_lookup, adversarial_review, follow_up, delegate). Descriptions are detailed and include usage guidance. However, there are gaps in output schema documentation (no explicit return type specifications in the visible schema definitions), and some parameter descriptions could be more precise about constraints. The tool-chaining model (chain arrays with fallback models) is well-designed, but there's no visible per-tool error handling guidance or recovery suggestions in the code. The server is designed as a smart dispatcher that routes queries to the agy CLI, which handles the heavy lifting, this is a clean composition pattern but means tool output depends entirely on agy's implementation (not visible here).
Get an adversarial second opinion from a different model family (Gemini Pro). ALWAYS use this for plan critiques, design reviews, and pre-merge code review: it hunts for flaws, edge cases, security issues, and unstated assumptions you may have missed.
Delegate file analysis to the Antigravity CLI (Gemini) instead of reading files yourself. USE THIS whenever a file is large (>200 lines) or the task spans more than 3 files: logs, database dumps, generated code, cross-file reviews, comparisons. The files never enter your context — only the answer does.
Delegate codebase archaeology to the Antigravity CLI: git log/diff/blame spelunking, wide greps across a repo, 'when/why did X change', 'where is Y used'. USE THIS instead of running many search commands yourself — it saves your context.
Raw delegation to the Antigravity CLI for heavy tasks that don't fit the other tools. agy has full tool access (shell, file reads, web) in the given cwd.
Continue a previous Antigravity session by session_id (returned by every other tool). USE THIS for follow-up questions about a prior delegation — the full prior context is already on agy's side, so you don't resend anything.
Output schemas not documented. None of the 6 tools declare their return type in visible Zod schemas. LLMs cannot plan downstream calls or extract specific fields without knowing the structure of responses.
No per-tool error handling or recovery guidance. The code does not show error responses, retry logic, or guidance for LLMs to self-correct (e.g. 'if session_id is invalid, start a new delegation'). Agents will not know whether failures are retryable.
Delegate tool is unconstrained. The 'delegate' tool accepts arbitrary prompts with no validation, length limits, or safety gates. It's marked as WRITE risk but has no dry-run, confirmation, or explicit destructiveness annotation. An agent could accidentally run destructive agy commands.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | C | 62 | 2026-07-28+ | v2 |
Delegate a web/documentation lookup to the Antigravity CLI (Gemini with web access): library docs, API references, error messages, current versions, external knowledge. USE THIS when you need information you don't have or that may be newer than your training data.
Mutual exclusivity not enforced in schema. adversarial_review's content and files parameters are both optional but mutually exclusive (validated in buildPrompt, not schema). Should use Zod's discriminatedUnion or oneOf to make this clear to LLMs.
No tool annotations. The visible code does not show toolAnnotations (readOnlyHint, destructiveHint, idempotentHint) in any tool definition. This omits useful classification for agents (especially critical for delegate which is marked WRITE but lacks destructiveHint).
Generic and vague naming for delegate tool. 'delegate' is too generic and doesn't convey what the tool does at a glance. Naming like 'delegate' requires the agent to read the full description to decide if it applies. Rename to 'execute_custom_task' or 'run_agy_prompt'.