gemini-vision-mcp
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
No arguments | |||
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": true
} |
| logging | {} |
| prompts | {
"listChanged": false
} |
| resources | {
"subscribe": false,
"listChanged": false
} |
| extensions | {
"io.modelcontextprotocol/ui": {}
} |
| experimental | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| vision_from_fileC | Describe a single file (image, audio, video, or text) using Gemini. |
| from_fileC | Alias for vision_from_file — describe a single file using Gemini. |
| vision_from_filesC | Describe multiple files in one Gemini session. |
| from_filesC | Alias for vision_from_files — describe multiple files using Gemini. |
| vision_from_files_mixedA | Describe multiple files of mixed types in one Gemini session. |
| from_files_mixedC | Alias for vision_from_files_mixed — describe mixed-type files using Gemini. |
| vision_from_resourceC | Describe a URL resource using Gemini. |
| from_resourceC | Alias for vision_from_resource — describe a URL using Gemini. |
| vision_from_resourcesC | Describe multiple URL resources in one Gemini session. |
| from_resourcesC | Alias for vision_from_resources — describe multiple URLs using Gemini. |
| vision_from_resources_mixedA | Describe multiple URL resources of mixed types in one Gemini session. |
| from_resources_mixedC | Alias for vision_from_resources_mixed — describe mixed-type URLs using Gemini. |
| new_visionA | Force a fresh Gemini Live session. Call this to reset the connection if it's stale or returns errors. Existing conversation history is lost. |
| gemini_usageA | Get current Gemini API usage statistics. Returns token usage stats including input/output tokens and rate limit status. |
| compactA | Compact conversation history and reconnect. Summarizes older turns, keeps recent ones verbatim, and replays into a fresh Gemini Live session so context carries over. Use instead of new_vision when you want to preserve context. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
| server_info | Return server configuration. |
TDQS
Scored across 15 tools
The 10 vision/from_* tools are essentially duplicate or near-duplicate aliases of each other (vision_from_file vs from_file, vision_from_files vs from_files, etc.), creating massive ambiguity. An agent cannot reliably distinguish which tool to pick when 12 of 15 tools describe the same core operation of 'describe files/resources with Gemini'.
There is a somewhat consistent pattern within the vision_from_* family (file/files/files_mixed, resource/resources/resources_mixed), but the alias tools (from_file, from_files, etc.) and the duplicate prefix 'vision_from_' broken by short aliases create two inconsistent naming conventions. The mix of descriptive prefixes with abbreviated aliases, plus session management tools (new_vision, compact) that don't follow the pattern, makes naming inconsistent overall.
15 tools is within the borderline range, but roughly two-thirds of them are redundant aliases of the same 5-6 core operations. The effective distinct tool surface is only about 5 unique operations inflated to 15, making the count feel padded rather than well-scoped.
The core input variants (single vs multiple, file vs resource, mixed types) are covered, and session management (reset, compact, usage stats) is present. However, there are no tools for more advanced capabilities like providing prompts/instructions to the vision model, and the heavy alias duplication obscures whether genuinely useful operations are missing.