Search: llm
Search MCP servers and agent skills by name, description, category or topic — 553 results.
NVIDIA/TensorRT-LLM/ad-accuracy-debug
Debug AutoDeploy accuracy regressions vs a reference score (PyTorch backend or published baseline).
NVIDIA/TensorRT-LLM/ad-add-fusion-transformation
Claude Code skill (trtllm-agent-toolkit): implement or extend TensorRT-LLM AutoDeploy fusion transforms under transform/library/ in a TensorRT-LLM checkout.
NVIDIA/TensorRT-LLM/ad-conf-check
Check whether AutoDeploy YAML configs were actually applied by analyzing server logs and optionally graph dumps (AD_DUMP_GRAPHS_DIR).
NVIDIA/TensorRT-LLM/ad-graph-dump
Enable and interpret TensorRT-LLM AutoDeploy FX graph text dumps via AD_DUMP_GRAPHS_DIR.
NVIDIA/TensorRT-LLM/ad-layer-visualizer
Visualize a specific transformer decoder layer from an AutoDeploy FX graph text dump as a hierarchical DOT/PNG diagram.
NVIDIA/TensorRT-LLM/ad-model-onboard
Translates a HuggingFace model into a prefill-only AutoDeploy custom model using reference custom ops, validates with hierarchical equivalence tests.
NVIDIA/TensorRT-LLM/exec-local-compile
Compile TensorRT-LLM on a compute node inside a Docker container.
NVIDIA/TensorRT-LLM/kernel-cute-writing
Write and implement GPU kernels using NVIDIA CuTe DSL (CUTLASS 4.x Python API) — NOT for Triton, CUDA C++, or conceptual explanations.
NVIDIA/TensorRT-LLM/kernel-tileir-optimization
Optimize existing Triton kernels for NVIDIA TileIR backend on Blackwell GPUs (sm_100+).
NVIDIA/TensorRT-LLM/kernel-triton-writing
ONLY for OpenAI Triton (@triton.jit) kernel development.
NVIDIA/TensorRT-LLM/perf-host-analysis
Analyze host/CPU overhead in TensorRT-LLM inference from nsys traces.
NVIDIA/TensorRT-LLM/perf-host-optimization
Profiles and optimizes TensorRT-LLM host/CPU overhead using line_profiler (with nsys support planned).
NVIDIA/TensorRT-LLM/perf-nsight-compute-analysis
Analyze ncu (NVIDIA Nsight Compute) profiling output: SOL% bottleneck classification, roofline analysis, occupancy diagnosis, memory hierarchy analysis, warp stall analysis, metr...
NVIDIA/TensorRT-LLM/perf-nsight-systems
Nsight Systems (nsys) CLI for system-level timeline profiling.
NVIDIA/TensorRT-LLM/perf-optimization
Performance optimization coordination playbook.
NVIDIA/TensorRT-LLM/perf-torch-cuda-graphs
Apply CUDA Graphs to PyTorch workloads — API selection (torch.compile, PyTorch make_graphed_callables, TE make_graphed_callables, MCore CudaGraphManager, FullCudaGraphWrapper, m...
NVIDIA/TensorRT-LLM/perf-torch-sync-free
Identify and eliminate host-device synchronizations in PyTorch code.
NVIDIA/TensorRT-LLM/perf-workload-profiling
Code instrumentation for timing workloads.
NVIDIA/TensorRT-LLM/trtllm-code-contribution
Best practices for contributing code to TensorRT-LLM.
NVIDIA/TensorRT-LLM/trtllm-codebase-exploration
Systematic approach to exploring the TensorRT-LLM codebase before implementing new features or optimizations.
NVIDIA/TensorRT-LLM/trtllm-flashinfer-upgrade
Upgrade flashinfer-python version in TensorRT-LLM.
NVIDIA/TensorRT-LLM/trtllm-moe-develop
Review, design, and refactor TensorRT-LLM PyTorch MoE code for architecture fit, clean code, maintainability, and testability.
NVIDIA/TensorRT-LLM/trtllm-serve-config-guide
Generate a source-backed starting `trtllm-serve --config` YAML for basic aggregate single-node PyTorch serving, aligned with checked-in TensorRT-LLM configs and deployment docs.
MarcellM01/TinySearch
Self-hosted web research for MCP agents: search (SearXNG, with DuckDuckGo fallback), crawl, dense+BM25 rerank, and dedupe into a source-grounded, cited prompt. Local ONNX embeddings by default, or bring an OpenAI-compatible embedding API.
ypollak2/llm-router
Subscription-aware LLM router for Claude Code. Routes tasks to 20+ providers (OpenAI, Gemini, Groq, Ollama, Codex) based on complexity classification, Claude subscription pressure, and cost. Free tasks stay on Claude subscription; expensive tasks fall back to the cheapest capable model. Includes 30 MCP tools, 6 auto-routing hooks, semantic dedup cache, prompt caching, daily spend cap, and a live w
smigolsmigol/llmkit
AI API cost tracking and budget enforcement across 11 LLM providers. 6 tools for spend analytics, budget monitoring, session summaries, and key management.
sametbrr/llm-wiki-manager
Persistent LLM-managed personal wiki — the model writes, cross-references, and maintains the knowledge base while you curate sources. Implements Karpathy's LLM Wiki pattern with 8 operating modes.
LLM-Pulse/llmpulse-mcp
Hosted MCP server for LLM Pulse AI visibility analytics, including brand mentions, citations, sentiment, share of voice, recommendations, GEO Writer, Search Console, and AI traffic via a Streamable HTTP endpoint.
ellmos-ai/ellmos-filecommander-mcp
Comprehensive local filesystem MCP server with file management, process control, interactive sessions, async search, and safe delete via Recycle Bin.
ellmos-ai/ellmos-codecommander-mcp
Developer-focused MCP server for code analysis, JSON repair, encoding fixes, and import organization across local projects.
zhaoyue722/llm-usage-mcp
Local-first LLM API cost tracker. Captures usage across Anthropic, OpenAI, Qwen, and DeepSeek into a local SQLite ledger and exposes spend queries, provider comparison, and recommendations as MCP tools — with first-class Chinese-provider support (CNY→USD). Install: `uvx llm-usage-mcp`.
ellmos-ai/n8n-manager-mcp
MCP server for managing n8n workflows through AI assistants, including workflow CRUD, synchronization, inspection, and execution support.
danieldoderlein/llm-bus
Multi-agent coordination bus over MCP: atomic gap-free claims, advisory file leases, a shared event ledger, presence, prose handoffs, and a task graph, so multiple coding agents (Claude Code, Cursor, Codex) stop colliding and re-deriving each other's work. Remote Streamable HTTP; self-hostable (AGPL-3.0) or hosted at llm-bus.com.
MGM-FALCON/quelllm-mcp
MCP server for [quelllm.fr](https://quelllm.fr) — open-weights LLM catalog (250+ models with params, VRAM by quantization, tokens/sec, license). Tools: `list_models`, `get_model`, `compare`, `estimate_vram`, `estimate_cost` (API providers vs self-hosted hardware in EUR), `search_models`. Useful when picking a local model that fits your GPU/Mac.
Jwrede/llmprobe
Synthetic monitoring for LLM inference endpoints. Measure TTFT, latency, throughput, and errors across OpenAI, Anthropic, Google, Azure, Bedrock, and local servers (vLLM, SGLang, Ollama). CLI + MCP server with Prometheus and OpenTelemetry export.
hivemindunit/llmintel-mcp
Check whether an LLM model id is deprecated, retiring, or retired, and what to migrate to. Covers OpenAI, Anthropic, Azure, Bedrock, Google, and Cohere. No API key. `npx -y @llmintel/mcp` or `https://llmintel.ai/v1/mcp`
romanshumy/llm-prices-data
Live LLM API pricing from modelpricewatch.com — current token prices, model comparisons, cheapest-model lookups, and The LLM Price Index across 150+ models from 20+ providers, re-verified daily against official provider pages. No API key required. `npx -y @modelpricewatch/mcp`
Daichi-Kudo/llm-advisor-mcp
Real-time LLM/VLM model comparison with benchmarks, pricing, and personalized recommendations from 5 data sources. No API key required.
JamesANZ/cross-llm-mcp
An MCP server that enables cross-LLM communication and memory sharing, allowing different AI models to collaborate and share context across conversations.
szp2005/llm-prices-cn
Daily-verified LLM API pricing dataset (44+ models, CN & global) with a hosted MCP server for live price queries and token cost estimation.
llmops-pro/marketplace-mcp
Run a Shopstr / NOSTR marketplace storefront from an agent: create and update stalls and products in both the NIP-15 dialect (kind 30017/30018) and the Shopstr-modern dialect (kind 30019 shop + kind 30402 NIP-99 listings), including the Shopstr cache POST that makes cards actually render. 7 tools. `npx -y marketplace-mcp`.
0xshellming/mcp-summarizer
AI Summarization MCP Server, Support for multiple content types: Plain text, Web pages, PDF documents, EPUB books, HTML content
wn01011/llm-token-tracker
Token usage tracker for OpenAI and Claude APIs with MCP support, real-time session tracking, and accurate pricing for 2025 models
elhamid/llm-council
Multi-LLM deliberation with anonymized peer review. Runs a 3-stage council: parallel responses → anonymous ranking → synthesis. Based on Andrej Karpathy's LLM Council concept.
microsoft/markitdown
MCP tool access to MarkItDown -- a library that converts many file formats (local or remote) to Markdown for LLM consumption.