@chrischall/gemini-mcp 2.1.3 → 2.2.0

package/dist/tools/library.js

@@ -3,7 +3,7 @@ import { McpToolError, minifiedResult } from '@chrischall/mcp-utils';
 import { LIBRARY_NAME_PATTERN } from '../library.js';
 import { acceptableUploadType } from '../upload-url.js';
 import { withProgressHeartbeat } from './shared.js';
-import { previewUnlessConfirmed, schemaConfirm } from './_confirm.js';
+import { confirmNote, confirmTokenParam, confirmWrite } from './_confirm.js';
 /**
  * Tools over the per-account reference library (src/library.ts): named
  * characters and styles that persist across sessions with NO expiry.
@@ -111,11 +111,21 @@ export function registerLibraryTools(server, client) {
         });
     });
     server.registerTool('gemini_delete_character', {
-        description: 'Delete a saved character from this account\'s reference library. Generation calls naming it will then fail until it is re-saved.',
+        description: 'Delete a saved character from this account\'s reference library. Generation calls naming it will then fail until it is re-saved. ' + confirmNote(),
         annotations: { readOnlyHint: false, destructiveHint: true, openWorldHint: false },
-        inputSchema: z.object({ name: NAME_FIELD, confirm: schemaConfirm }),
-    }, async (args) => {
-        const gate = previewUnlessConfirmed(args.confirm, 'Delete a saved character (its reference image and description)', 'DELETE', `characters/${args.name}`);
+        inputSchema: z.object({ name: NAME_FIELD, confirmToken: confirmTokenParam }),
+    }, async (args, extra) => {
+        const gate = await confirmWrite(extra, {
+            tool: 'gemini_delete_character',
+            action: 'character.delete',
+            message: 'Review and confirm this deletion:',
+            description: 'Delete a saved character (its reference image and description)',
+            method: 'DELETE',
+            path: `characters/${args.name}`,
+            target: args.name,
+            payload: { name: args.name },
+            confirmToken: args.confirmToken,
+        });
         if (gate)
             return gate;
         const existed = await library.deleteCharacter(args.name);
@@ -170,11 +180,21 @@ export function registerLibraryTools(server, client) {
         });
     });
     server.registerTool('gemini_delete_style', {
-        description: 'Delete a saved style preset from this account\'s reference library.',
+        description: 'Delete a saved style preset from this account\'s reference library. ' + confirmNote(),
         annotations: { readOnlyHint: false, destructiveHint: true, openWorldHint: false },
-        inputSchema: z.object({ name: NAME_FIELD, confirm: schemaConfirm }),
-    }, async (args) => {
-        const gate = previewUnlessConfirmed(args.confirm, 'Delete a saved style preset', 'DELETE', `styles/${args.name}`);
+        inputSchema: z.object({ name: NAME_FIELD, confirmToken: confirmTokenParam }),
+    }, async (args, extra) => {
+        const gate = await confirmWrite(extra, {
+            tool: 'gemini_delete_style',
+            action: 'style.delete',
+            message: 'Review and confirm this deletion:',
+            description: 'Delete a saved style preset',
+            method: 'DELETE',
+            path: `styles/${args.name}`,
+            target: args.name,
+            payload: { name: args.name },
+            confirmToken: args.confirmToken,
+        });
         if (gate)
             return gate;
         const existed = await library.deleteStyle(args.name);

package/dist/tools/music.js

@@ -5,7 +5,7 @@ import { DEFAULT_MUSIC_MODEL } from '../models.js';
 import { emitMedia, timeoutMsSchema, idempotencyKeySchema, asyncSchema, maxWaitMsSchema, withProgressHeartbeat, assertLocalInputsAvailable, imagesUrlSchema, imagesFileUrisSchema, imagesBase64Schema } from './shared.js';
 import { fingerprintRequest } from '../jobs.js';
 import { attachCost } from '../pricing.js';
-import { previewLocalInputsUnlessConfirmed, schemaConfirm } from './_confirm.js';
+import { confirmLocalInputs, confirmNote, confirmTokenParam } from './_confirm.js';
 const MUSIC_MODELS = ['lyria-3-clip-preview', 'lyria-3.5', 'lyria-3-pro-preview'];
 /**
  * There is no output-format parameter, and there is no chaining.
@@ -28,7 +28,8 @@ export function registerMusicTools(server, client) {
             'lyria-3-clip-preview (30s instrumental clip, default, cheapest), lyria-3.5 (full-length song with vocals) or ' +
             'lyria-3-pro-preview (longer-form). Output is MP3, written to disk (or returned inline). Single-turn: a track ' +
             'cannot be refined by a follow-up call, so put the whole brief in the prompt. Runs long — give it a `max_wait_ms` ' +
-            'budget (or `async: true` + gemini_get_result on a local install), or raise `timeout_ms`. Needs a funded account.',
+            'budget (or `async: true` + gemini_get_result on a local install), or raise `timeout_ms`. Needs a funded account. ' +
+            confirmNote('Local file inputs are confirmed first'),
         annotations: { readOnlyHint: false, openWorldHint: true },
         inputSchema: z.object({
             prompt: z.string().min(1).describe('Description of the music: mood, genre, instruments, tempo, structure, or lyrics'),
@@ -46,12 +47,16 @@ export function registerMusicTools(server, client) {
             idempotency_key: idempotencyKeySchema,
             async: asyncSchema,
             max_wait_ms: maxWaitMsSchema,
-            confirm: schemaConfirm,
+            confirmToken: confirmTokenParam,
         }),
     }, async (args, extra) => {
         assertLocalInputsAvailable(client.mediaSink, args);
         const model = args.model ?? DEFAULT_MUSIC_MODEL;
-        const gate = await previewLocalInputsUnlessConfirmed(args.confirm, 'Send local image input(s) to the Gemini Lyria API', '/v1beta/interactions', args.images);
+        const { confirmToken, ...request } = args;
+        const gate = await confirmLocalInputs(extra, {
+            tool: 'gemini_music_generate', action: 'music.generate', description: 'Send local image input(s) to the Gemini Lyria API', endpoint: '/v1beta/interactions',
+            imagePaths: args.images, request, confirmToken,
+        });
         if (gate)
             return gate;
         const fingerprint = fingerprintRequest('gemini_music_generate', {

package/dist/tools/set.js

@@ -7,11 +7,12 @@ import { emit, resolveAspectRatio, ASPECT_RATIOS, sharedImageSchema, pickSeed, b
 import { fingerprintRequest } from '../jobs.js';
 import { sumUsage } from '../usage.js';
 import { attachCost } from '../pricing.js';
-import { previewLocalInputsUnlessConfirmed, schemaConfirm } from './_confirm.js';
+import { confirmLocalInputs, confirmNote, confirmTokenParam } from './_confirm.js';
 export function registerSetTools(server, client) {
     server.registerTool('gemini_image_set', {
         description: 'Generate a consistent SET of images: a master image from master_prompt, then one image per scene that references the master so the subject/style stays consistent. Provide `scenes` (explicit per-image prompts) OR `count` (variations of the master). ' +
-            'Scene generations run in parallel (reference_mode "master", the default). On the hosted connector: saved `characters` and a saved `style` can seed the whole set by name, multi-image results include a `bundle_url` zip of every image (one curl instead of N), and `max_wait_ms` returns a pollable job handle if the batch runs long.',
+            'Scene generations run in parallel (reference_mode "master", the default). On the hosted connector: saved `characters` and a saved `style` can seed the whole set by name, multi-image results include a `bundle_url` zip of every image (one curl instead of N), and `max_wait_ms` returns a pollable job handle if the batch runs long. ' +
+            confirmNote('Local file inputs are confirmed first'),
         annotations: { readOnlyHint: false, openWorldHint: true },
         inputSchema: z.object({
             master_prompt: z.string().min(1).describe('Prompt for the master/reference image'),
@@ -26,7 +27,7 @@ export function registerSetTools(server, client) {
             characters: charactersSchema,
             style: styleSchema,
             master_images_base64: imagesBase64Schema('Reference images for the master generation', 'master_images'),
-            confirm: schemaConfirm,
+            confirmToken: confirmTokenParam,
             ...sharedImageSchema,
         }),
     }, async (args, extra) => {
@@ -36,8 +37,12 @@ export function registerSetTools(server, client) {
             throw new McpToolError('Provide exactly one of `scenes` or `count`.');
         }
         // Confirm-gate local file inputs: only `master_images` (paths) can read a
-        // local file; base64 references pass through ungated. Dry-run makes NO API call.
-        const gate = await previewLocalInputsUnlessConfirmed(args.confirm, 'Send local master image input(s) to the Gemini API', '/v1beta/models/{model}:generateContent', args.master_images);
+        // local file; base64 references pass through ungated. The preview makes NO API call.
+        const { confirmToken, ...request } = args;
+        const gate = await confirmLocalInputs(extra, {
+            tool: 'gemini_image_set', action: 'image.set', description: 'Send local master image input(s) to the Gemini API', endpoint: '/v1beta/models/{model}:generateContent',
+            imagePaths: args.master_images, request, confirmToken,
+        });
         if (gate)
             return gate;
         const model = resolveModel(args.model, readEnvVar('GEMINI_IMAGE_MODEL'));

package/dist/tools/video.js

@@ -6,7 +6,7 @@ import { DEFAULT_VIDEO_MODEL } from '../models.js';
 import { emitMedia, reportShape, resolveAspectRatio, orientationSchema, timeoutMsSchema, idempotencyKeySchema, asyncSchema, maxWaitMsSchema, withProgressHeartbeat, assertLocalInputsAvailable, imagesUrlSchema, imagesFileUrisSchema, imagesBase64Schema } from './shared.js';
 import { fingerprintRequest } from '../jobs.js';
 import { attachCost } from '../pricing.js';
-import { previewLocalInputsUnlessConfirmed, schemaConfirm } from './_confirm.js';
+import { confirmLocalInputs, confirmNote, confirmTokenParam } from './_confirm.js';
 /** omni's own aspect-ratio enum — NOT the image tools' ASPECT_RATIOS. */
 const VIDEO_ASPECT_RATIOS = ['16:9', '9:16'];
 const VIDEO_TASKS = ['text_to_video', 'image_to_video', 'reference_to_video', 'edit', 'extend'];
@@ -36,7 +36,8 @@ export function registerVideoTools(server, client) {
             'or continue a prior video (task: "edit" or "extend" + previous_interaction_id / continue_last; extensions ' +
             'add ~3-10s each, to ~40s total). Cost scales with `resolution` — draft at 360p, keep at 1080p/4k. ' +
             'Output is written to disk as MP4 (video has no inline MCP block). Video runs long — give it a `max_wait_ms` budget ' +
-            '(or `async: true` + gemini_get_result on a local install), or raise `timeout_ms`. Needs a funded account.',
+            '(or `async: true` + gemini_get_result on a local install), or raise `timeout_ms`. Needs a funded account. ' +
+            confirmNote('Local file inputs are confirmed first'),
         annotations: { readOnlyHint: false, openWorldHint: true },
         inputSchema: z.object({
             prompt: z.string().min(1).describe('Description of the video to generate (or the edit instruction when task=edit)'),
@@ -64,7 +65,7 @@ export function registerVideoTools(server, client) {
             idempotency_key: idempotencyKeySchema,
             async: asyncSchema,
             max_wait_ms: maxWaitMsSchema,
-            confirm: schemaConfirm,
+            confirmToken: confirmTokenParam,
         }),
     }, async (args, extra) => {
         assertLocalInputsAvailable(client.mediaSink, args);
@@ -79,7 +80,11 @@ export function registerVideoTools(server, client) {
         }
         // Confirm-gate local file inputs (a prompt-injected `images` path could
         // exfiltrate a local file); base64 / clipboard pass through ungated.
-        const gate = await previewLocalInputsUnlessConfirmed(args.confirm, 'Send local image input(s) to the Gemini omni API', '/v1beta/interactions', args.images);
+        const { confirmToken, ...request } = args;
+        const gate = await confirmLocalInputs(extra, {
+            tool: 'gemini_video_generate', action: 'video.generate', description: 'Send local image input(s) to the Gemini omni API', endpoint: '/v1beta/interactions',
+            imagePaths: args.images, request: { ...request, previous_interaction_id: previousInteractionId }, confirmToken,
+        });
         if (gate)
             return gate;
         const model = args.model?.trim() || DEFAULT_VIDEO_MODEL;

package/dist/version.js

@@ -1,2 +1,2 @@
 /** Single source of the server version. release-please bumps the literal below. */
-export const VERSION = '2.1.3'; // x-release-please-version
+export const VERSION = '2.2.0'; // x-release-please-version

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@chrischall/gemini-mcp",
-  "version": "2.1.3",
+  "version": "2.2.0",
   "mcpName": "io.github.chrischall/gemini-mcp",
   "description": "Gemini image-generation MCP server for Claude — developed and maintained by AI (Claude Code)",
   "author": "Claude Code (AI) <https://www.anthropic.com/claude>",
@@ -46,7 +46,7 @@
     "typecheck": "tsc -p tsconfig.test.json"
   },
   "dependencies": {
-    "@chrischall/mcp-utils": "^2.4.0",
+    "@chrischall/mcp-utils": "^2.6.0",
     "@modelcontextprotocol/server": "^2.0.0",
     "dotenv": "^17.4.0",
     "zod": "^4.6.5"

package/server.json

@@ -6,12 +6,12 @@
     "url": "https://github.com/chrischall/gemini-mcp",
     "source": "github"
   },
-  "version": "2.1.3",
+  "version": "2.2.0",
   "packages": [
     {
       "registryType": "npm",
       "identifier": "@chrischall/gemini-mcp",
-      "version": "2.1.3",
+      "version": "2.2.0",
       "transport": {
         "type": "stdio"
       },

package/skills/gemini-mcp/SKILL.md

@@ -125,9 +125,9 @@ not an error.
 ### Files API
 | Tool | Description |
 |------|-------------|
-| `gemini_upload_file(url? \| data_base64? \| path?, mime_type?, display_name?, confirm?)` | Upload once, get a reusable `files/<id>`. Exactly one source. `url` is fetched by the server (image/video/audio, ≤100MB); `path` is stdio-only and confirm-gated; `data_base64` is the last resort |
+| `gemini_upload_file(url? \| data_base64? \| path?, mime_type?, display_name?, confirmToken?)` | Upload once, get a reusable `files/<id>`. Exactly one source. `url` is fetched by the server (image/video/audio, ≤100MB); `path` is stdio-only and confirmed first (see Notes); `data_base64` is the last resort |
 | `gemini_list_files(page_size?)` | List current uploads with MIME types and expiry |
-| `gemini_delete_file(file_uri, confirm)` | Delete an upload before its ~48h expiry |
+| `gemini_delete_file(file_uri, confirmToken?)` | Delete an upload before its ~48h expiry (confirmed first — see Notes) |
 
 On the **hosted hosted deployment** there is also `POST /upload`, behind the same OAuth token as
 `/mcp` — the zero-base64 path for an agent with a shell:
@@ -307,6 +307,7 @@ Condensed from Google's official Nano Banana prompting guide. Core rule:
 
 ## Notes
 
+- **Confirmations.** Local file inputs (`images`, `master_images`, `video_path`, `gemini_upload_file`'s `path`) and every delete are confirmed before anything is sent: a confirmation prompt where the client supports one; otherwise the first call does nothing and returns `status: "confirmation-required"` with a `preview` (resolved paths, MIME types, sizes — or the method/path being deleted) and a `confirmToken`. Show the preview to the user, and only after they approve call again with the **same arguments** plus `confirmToken`. The token is single-use, expires (default 10 min), and is bound to those exact arguments — change anything and the call is refused (`DRAFT_CHANGED`, with a fresh preview and token). `MCP_CONFIRM_MODE` on the server picks `ask-user` (default), `auto` (the model may approve after reviewing the preview) or `refuse`. Text prompts, URLs, `files/` references and base64 inputs are not gated.
 - **Input images** accept either file **paths** (`images` / `master_images`) or **base64/data-URI values** (`images_base64` / `master_images_base64`).
 - **`seed`** makes a result reproducible; it's echoed in the result metadata (a random one is chosen + echoed when omitted). `count>1` uses `seed, seed+1, …` so the images differ. Determinism isn't fully guaranteed by the model.
 - **`filename`/`basename`** set the output name (extension stripped); names never overwrite (a `-2`, `-3` suffix is added). The result echoes the absolute path(s), `model`, `seed`, and aspect/size.