mcp-nvidia-nim
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| NVIDIA_API_KEY | Yes | NVIDIA NIM API Key (nvapi-... from build.nvidia.com) | |
| NVIDIA_BASE_URL | No | Base URL (official endpoint) | https://integrate.api.nvidia.com/v1 |
| NVIDIA_DEFAULT_MODEL | No | Default model for standard queries | mistralai/mistral-large-4-1t-a52b |
| NVIDIA_RATE_LIMIT_RPM | No | Rate limit per minute (standard free tier is 40 RPM) | 40 |
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
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| nim_query_modelB | Executes a direct prompt query against any NVIDIA NIM model (e.g. Nemotron Ultra 550B, Mistral Large, MiniMax Flash, etc.) |
| nim_deep_reasonA | Leverages NVIDIA large-scale reasoning models (Nemotron-3 Ultra 550B or Nemotron-3 Omni Reasoning) for deep system architecture, complex logic, and tradeoff analysis |
| nim_review_codeC | Audits source code or git diffs using software engineering specialist models (Poolside Laguna XS or Mistral Large) |
| nim_analyze_visionC | Analyzes images, architecture diagrams, screenshots, and wireframes using large multimodal models (Llama 3.2 90B Vision or Nemotron Super VL 120B) |
| nim_extract_documentB | Processes and structures raw dumps from PDFs, tables, unorganized notes, or OCR using Nemotron Parse 2.0 |
| nim_statusA | Reports rate limiting status (40 RPM quota), available slots, execution telemetry, and supported model catalog and aliases |
| nim_consultar_modeloB | Alias para nim_query_model (compatibilidade com versões anteriores) |
| nim_raciocinar_profundoC | Alias para nim_deep_reason (compatibilidade com versões anteriores) |
| nim_revisar_codigoC | Alias para nim_review_code (compatibilidade com versões anteriores) |
| nim_analisar_visaoC | Alias para nim_analyze_vision (compatibilidade com versões anteriores) |
| nim_extrair_documentoC | Alias para nim_extract_document (compatibilidade com versões anteriores) |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 11 tools
Half the tool set consists of explicit aliases (Portuguese names) that duplicate the English tools exactly, making it impossible to infer any selection difference. The six unique capabilities are distinct, but the aliases create multiple unclear boundaries and invite misselection.
All names use a consistent nim_ prefix and snake_case, which is readable. However, the set mixes English and Portuguese conventions, and the alias pairs introduce duplicate naming patterns rather than a single predictable scheme.
Eleven tools is a reasonable raw count, but five are redundant aliases, leaving only six unique capabilities. The aliases inflate the surface without adding functionality, making the count feel heavier than necessary.
The unique tools cover the core NIM lifecycle: status/quota inspection, general query, deep reasoning, code review, vision analysis, and document extraction. Minor gaps exist (e.g., no explicit model listing beyond status), but agents can work around them.