Gemini Relay • Antigravityv1.2.0
MCP Tools Reference
gemini-relay exposes a comprehensive suite of tools over standard MCP protocol to empower AI coding assistants.
🛠️ Tool Catalog
1. ask-gemini
Primary intelligence and analysis endpoint. Handles questions, large codebase exploration, and refactoring suggestions.
| Parameter | Type | Required | Description |
|---|---|---|---|
prompt | string | Yes | Analysis request. Supports @src/file.ts syntax to inline project files. |
model | string | No | Gemini model (default: 'gemini-3.8-flash-high'). Also accepts 'gemini-3.1-pro-high', 'flash', 'pro'. |
effort | enum | No | "low" | "medium" | "high" — Allocates thinking token depth. |
mode | enum | No | "plan" (read-only) | "accept-edits" (automated editing). |
jsonSchema | object | string | No | Enforces structured JSON output conforming to this schema. |
addDirs | string[] | No | Additional workspace directories added to context. |
includeUsage | boolean | No | Default false. When true, returns thinking tokens and latency metrics. |
changeMode | boolean | No | Default false. Outputs structured <<<< OLD / NEW >>>> diff blocks. |
sandbox | boolean | No | Run in sandbox mode where supported. |
2. gemini-plan
Dedicated architectural planner using Gemini's plan mode (--mode plan) and maximum reasoning effort.
| Parameter | Type | Required | Description |
|---|---|---|---|
task | string | Yes | The architectural feature, refactor, or problem to design. |
context | string | No | Constraints, reference files (@file), or requirements. |
model | string | No | Default 'gemini-3.8-flash-high' or 'gemini-3.1-pro-high'. |
effort | enum | No | Default "high". Deep thinking token allocation. |
includeUsage | boolean | No | Default true. Returns token and latency metrics. |
3. gemini-image
Generates high-definition imagery and visual assets using Google Imagen & Gemini.
| Parameter | Type | Required | Description |
|---|---|---|---|
prompt | string | Yes | Detailed visual description (subject, lighting, composition, style). |
aspectRatio | enum | No | '1:1' (default), '16:9', '9:16', '4:3', '3:4', '3:2', '2:3'. |
outputPath | string | No | Relative project workspace path to automatically save the image (e.g. 'assets/hero.png'). |
4. gemini-models
Introspects available Gemini models, default selections, reasoning tiers, and active backend capabilities.
- Parameters: None (
{})
5. gemini-doctor
Diagnoses your environment, detecting installed CLIs (agy / gemini), versions, and system readiness.
- Parameters: None (
{})
6. brainstorm
Structured creative ideation engine utilizing established design frameworks (SCAMPER, Design Thinking, Lateral Thinking, Divergent/Convergent).
| Parameter | Type | Required | Description |
|---|---|---|---|
prompt | string | Yes | Core challenge or topic to explore. |
methodology | enum | No | 'divergent' | 'convergent' | 'scamper' | 'design-thinking' | 'lateral' | 'auto'. |
effort | enum | No | Reasoning effort level ("low", "medium", "high"). |
domain | string | No | Target domain (e.g. 'software', 'security'). |
ideaCount | number | No | Default 12. Number of concepts to generate. |
7. fetch-chunk
Retrieves subsequent paginated chunks from large multi-file edit suggestions generated during changeMode.
- Parameters:
cacheKey(string),chunkIndex(number).
8. ping & Help
ping: Verifies server connection and echo latency.Help: Retrieves the active CLI backend's official help manual.