MCP servers that ship as code — a repository or a package you install and run locally (usually over stdio). They have no hosted endpoint, so we cannot probe or grade them: instead each one shows its repository, install command, license, stars and when it was last updated.
Collected from GitHub repositories and npm, PyPI, Cargo, Go, Maven, NuGet and RubyGems packages tagged as MCP servers, the official MCP Registry, Docker's MCP catalog, ModelScope, Baidu MCP World, Cline and other public directories; one entry per repository. A server that also offers a hosted endpoint is listed under Remote instead. These entries are not counted in our reachability, verification or readiness statistics.
-
gosce-traced-llm-mcp-proxyupdated this year
Proxy Gemini (Vertex AI) completions wrapped in OpenTelemetry trace spans; returns the answer plus the trace/span id.
pypi MIT -
climategpt-performance-enhancementgithub.com/Mkatika37/climategpt-performance-enhancementupdated this year
Grounds a climate language model in real satellite (NASA VIIRS) and water-risk (WRI Aqueduct) data via MCP tools, so answers trace back to evidence, not just the model's memory.
-
@mtuska/ue-trace-mcpgithub.com/mtuska/ue-trace-mcpupdated this year
MCP server for LLM-driven Unreal Engine .utrace profiling. Wraps the headless TraceDigest UE Program and manages a long-lived daemon per trace.
npm Unlicense -
021-mcpupdated this year
MCP server for interacting with LLM traces - search, analyze, and debug your AI agent executions
pypi MIT -
MCP server for Arize Phoenix — LLM tracing, evaluation, and dataset management via AI agents
MIT -
iflow-mcp_traceloop-opentelemetry-mcp-servergithub.com/traceloop/opentelemetry-mcp-serverupdated this year
MCP server for querying OpenTelemetry traces from LLM applications with Opentelemetry support
pypi Apache-2.0 -
Self-hosted LLM & agent tracing with infinite retention: a Next.js dashboard on Vercel backed by your own Firebase project, with scoped API keys, a REST API, a TypeScript SDK, and an MCP server so AI agents can work with traces.
MIT -
Rust-native hybrid retrieval over LLM thinking traces - dense ANN + BM25 + rerank, all local ONNX. Daemon, HTTP + MCP, zero API keys.
Apache-2.0 ★ 1 -
Benchmark and MCP tools for LLM agents doing root-cause analysis on OpenTelemetry metrics, traces and logs. 20 recorded microservice incidents, every run scored against a 28-line rule and replayable. Only the strongest models clearly win.
Apache-2.0 -
@nexus2520/dep-tracer-mcp-servergithub.com/pdogra1299/dep-tracer-mcpupdated this year
MCP server that maps function and component dependencies in Haskell and ReScript codebases — built to help LLMs trace affected code flows for test generation.
npm MIT -
perfdigest-mcpgithub.com/onlyxitachi/perfdigest-mcpupdated this year
Local MCP server that digests development-loop reports — profilers (NVIDIA ncu, AMD rocprof, Linux perf, Apple Metal, chrome-trace), build tools (ptxas, clang -ftime-trace, cmake, ninja, cargo), benchmarks (criterion), CI run logs, and git change state — into token-efficient JSON for LLM coding agents.
pypi Apache-2.0 -
MCP server for agentic LLM evaluation: jury scoring, agent tracing via OpenTelemetry, document-grounded QA generation, PDF reports.
pypi Apache-2.0 -
systracesuitegithub.com/r1chard-lyu/systracesuiteupdated this year
Systracesuite is a MCP server designed to expose Linux tracing and debugging capabilities such as bpftrace, perf, ftrace, and syscall tracing to LLMs and autonomous agents.
Apache-2.0 ★ 2 -
Agent-native bi-temporal graph store: verified temporal operators as LLM tools + trace-grounded answer verification
pypi Apache-2.0 ★ 5 -
An agentic music librarian with a deterministic validator, a repair loop, and a measured eval suite. LLM tool use, RAG, tracing and MCP.
MIT -
Jira sprint metrics computed by code, reported by an LLM, guarded against invented numbers. MCP server, evals and tracing included.
MIT -
In-app debug traces and an LLM-friendly HTTP + MCP debug server for Apple-platform apps
MIT -
Self-hosted LLM operations gateway in one Go binary: OpenAI-compatible router, durable agent tracker, faithfulness evals + drift, traces, MCP server, console
Apache-2.0 ★ 2 -
Local-first text and LLM dataset curation studio for JSONL, SFT, DPO, conversations, agent traces, quality review, versioning, exports, MCP, and modular AI plugins.
MIT ★ 2 -
gotg-cligithub.com/pageton/gotg-cliupdated this year
Telegram MTProto debug and invoke CLI — call any TL method, trace API calls, and expose Telegram as an MCP server for LLM-driven interaction
★ 1 -
See every hop of a request through your local LLM system: a tracing server, a live flow map, a waterfall per request, and an MCP server for Claude Code and Cursor. One static Go binary.
Apache-2.0 ★ 1 -
LLM agent for business websites: answers from a knowledge base, captures leads · OpenTelemetry GenAI tracing, evals, MCP server · FastAPI · RU/EN/AR
-
MCP server for Spanlens — query LLM cost, anomalies, traces, and per-user analytics from inside Cursor, Continue, or Claude Desktop.
npm MIT -
LLM observability and evals in a single binary: traces, prompts, datasets, scores and review queues for LLM and agent apps. Self-hosted, SQLite inside, OpenTelemetry-native, MCP and CLI for coding agents.
go Apache-2.0 ★ 1 -
LLM-optimized MCP server for Datadog APM, logs, and metrics. Delivers structured, token-efficient trace and log insights designed specifically for AI-driven debugging—not just raw Datadog data.
MIT ★ 1 -
Open-source, model-agnostic control plane for LLM agents — tamper-evident, human-gated, replayable run bundles with full trace history, model I/O, and cost telemetry. TypeScript · React · MCP · OpenAPI. MIT, published on npm.
npm MIT -
Log Claude Code sessions to Opik, the open-source LLM observability and evaluation platform, built by Comet. Tracing, evaluation, and skills for observable AI applications.
go Apache-2.0 ★ 25 -
LLM-native code toolkit: Rust multi-language parser (tree-sitter) + 44 MCP tools for atomic editing, impact analysis, reference tracing and deterministic zero-LLM code quality gates. Built for the handless, eyeless, memoryless LLM.
npm MIT ★ 1 -
CLI and MCP server that traces which attention heads and neurons drive an LLM's output, built on TransformerLens.
pypi MIT ★ 1 -
Unreal-Insights-MCP-Hivegithub.com/Camreyn/Unreal-Insights-MCP-Hiveupdated this year
MCP aggregator for Unreal Insights .utrace files that uses official UE TraceServices to extract, index, and expose per-trace query servers for LLM and MCP clients.
★ 2 -
Runtime debugging for LLM agents: launch, trace, break and observe live programs over MCP, no recompilation.
MIT -
Official Model Context Protocol (MCP) server for ModelGate: run LLM inference through the ModelGate gateway and inspect models, usage, cost and request traces from Claude, Cursor, VS Code and any MCP client.
npm MIT -
Evidence-grounded IT support agent: Gemini function calling over an MCP-style tool server (KB search, ticket search, ticket drafts), with a 60-question eval harness, LLM judge, and OpenTelemetry tracing.
★ 1 -
mcp-owlgithub.com/lucianopalhares/mcp-owlupdated this year
MCP server for codebase navigation — explains symbols, traces execution flows, and searches references via AST parsing and LLM. Supports PHP and TypeScript projects.
npm MIT -
Point it at any .NET repo and get a map you can trust — every edge cites file and line. Roslyn code graph with Map + Trace views, LLM context packs, MCP server. Tauri + Angular.
MIT ★ 1 -
Reference multi-agent banking assistant for UK retail banking: PII redaction before the LLM, MCP tool servers, customer-confirmed actions, cost and traces per turn.
-
MCP server for querying OpenTelemetry traces across multiple observability backends (Jaeger, Tempo, Traceloop, Datadog, and more) for LLM application debugging. Fork of traceloop/opentelemetry-mcp-server.
pypi Apache-2.0 ★ 1 -
HomelabSentinel — an AI SRE for my homelab that asks before it acts: LangGraph agent with a human-approval gate, MCP server, BM25 RAG, evals and Langfuse tracing on a 100% local LLM.
MIT -
RAG + LLM maintenance copilot: LangGraph multi-agent supervisor with human approval, hybrid search over equipment manuals (Qdrant, BM25, reranking), MCP server, sensor and image analysis, evaluation suite, Langfuse tracing. Runs locally with Ollama. FastAPI, Postgres.
MIT -
Open-source, self-hosted observability for AI agents: OTLP ingestion, tracing SDK, sessions, cost analytics, datasets and LLM-as-judge experiments, alerts, replay, and an MCP server. SQLite or PostgreSQL.
pypi MIT -
@hasna/microservicesgithub.com/hasna/microservicesupdated this year
21 production-grade microservice building blocks for AI-native SaaS — auth, billing, LLM gateway, agent registry, RAG, guardrails, tracing, and more. Each with PostgreSQL, HTTP API, MCP server, and CLI.
npm Apache-2.0 -
Voice + chat assistant with long-term memory built from scratch: LLM fact extraction, pgvector, conflict resolution (90% held-out accuracy vs 0% baseline), relevance-filtered RAG, a memory-poisoning screen, and Langfuse tracing. LangGraph, FastAPI, React/TS, MCP
-
Async RAG backend for PDF Q&A: FastAPI + Postgres/pgvector, streamed answers, swappable LLM providers (Groq, Bedrock, OpenAI, Azure), MCP search tool, Langfuse tracing, 38-question eval harness, pytest + GitHub Actions CI.
-
cowork-qa-mcpgithub.com/inSideos-designs/cowork-qa-mcpupdated this year
MCP server that runs goal-driven Playwright browser sessions with full action traces and aria-snapshots, so the calling LLM can verify whether a goal was met.
npm MIT -
A production-style multi-agent system that generates context-aware briefs. Built on MCP + A2A, with persistent memory (Chroma), dynamic tool routing, self-critique, distributed tracing (Arize Phoenix), an LLM-as-judge eval harness, and Redis caching.
★ 1 -
Multi-tenant operational intelligence platform AI control plane in Go/K8s. Features custom operator, Kafka ingestion, Temporal workflows, MCP LLM gateway, pgvector hybrid search, OpenTelemetry distributed tracing, ArgoCD GitOps, and Redis semantic caching
Apache-2.0 ★ 3 -
mcp-server-customgithub.com/themoizqureshi/mcp-server-customupdated this year
Production MCP server: 3 async tools (job search, salary lookup, LLM resume analysis) for Claude Desktop via stdio. Pydantic structured outputs, aiosqlite, LangSmith tracing.
-
Logs, traces and LLM calls from every run, live in VS Code: terminals, debug sessions, Docker and OpenTelemetry, next to production in Datadog, CloudWatch, Loki, Elasticsearch, Cloud Logging and SQL databases. Shared with Copilot, Claude Code or Codex over MCP.
MIT ★ 2 -
Three-agent research system on the OpenAI Agents SDK — Planner/Executor/Reviewer with typed handoffs, a custom MCP server, function tools, input/output guardrails, token-streaming (CLI + SSE), tracing, and an LLM-judge eval harness with a pytest regression gate.
-
Multimodal food concierge (in development): agents recommend dishes and plan meals from text or a photo, with diet and allergen limits enforced by code, not the LLM. MCP tools, hybrid retrieval, Langfuse tracing and evaluation, free-tier and local models only.
Apache-2.0 ★ 1 -
RAGv2-Agentic-RAGgithub.com/tumasv1/RAGv2-Agentic-RAGupdated this year
Agentic RAG assistant for a personal Obsidian vault — LangGraph ReAct agent, hybrid search (BM25+dense+reranker), MCP tools, LLM gateway with fallback, Langfuse tracing.
MIT -
A private AI finance copilot for investors: ask about your portfolio in plain language, get figures worked out in code with a trace of every answer, and keep your history locked with your own PIN. FastAPI, LangGraph and a local LLM via Ollama.
-
Deterministic web-evidence engine: trace one field across raw HTML, structured data, embedded state, rendered DOM and network JSON, with exact provenance. CLI and MCP. No LLM.
Apache-2.0 -
Correctness-enforcing double-entry ledger runtime for LLM agents that move money: deterministic core, durable command journal, policy and signed approvals, verifiable trace schema, stdio MCP server, OpenTelemetry GenAI adapter, red-team corpus with conformance levels and drift, mutation-gated. Pre-alpha (M0-M7 done), building in public.
NOASSERTION -
zeno-mobile-runnergithub.com/johnmikel/zeno-mobile-runnerupdated this year
Deterministic mobile UI checks for AI coding agents. One Zig binary drives real iOS/Android devices, returns a typed pass/fail, and writes a replayable trace you can gate a merge on. No LLM inside.
MIT ★ 5 -
A lean, local knowledge graph that joins a repo's code to its aSPARK delivery artifacts (specs, plans, reviews, QA) — served over MCP and a CLI. Trace which code implements a user story, and see the story-level blast radius of a change. Deterministic, offline, no LLM.
MIT ★ 3 -
@iflow-mcp/dcostenco-prism-mcp-servergithub.com/dcostenco/prism-mcpupdated this year
The Mind Palace for AI Agents — persistent memory (SQLite/Supabase), behavioral learning & IDE rules sync, multimodal VLM image captioning, pluggable LLM providers (OpenAI/Anthropic/Gemini/Ollama), OpenTelemetry distributed tracing, GDPR export, multi-age
npm MIT -
docwise-agenticgithub.com/rakmohan/docwise-agenticupdated this year
Full agentic RAG pipeline: adaptive chunking → Qdrant hybrid search (semantic + BM25 + RRF) → LangGraph ReAct agent → grounded answers with citations, streamed live. Multi-turn conversation, multi-hop reasoning, human-in-the-loop, LangSmith tracing, MCP server. LLM-agnostic — works with Claude or OpenAI.
MIT -
R&D tax-credit evidence from your git history: a tax year's commits by component and person, and the commits that look like experimentation (reverts, unmerged branches, spikes, benchmarks), each traced to its commit, in a dated pack for your accountant. Local; no account, no LLM. Not tax advice.
MIT -
Mirror your body in data. One standard, every source, one complete picture of you: a self-hosted health record built from lab reports, wearables and genetics, with every number traced to its source and a built-in MCP server. Run it with Docker and one LLM key, or keep every model local on llama.cpp.
pypi Apache-2.0 ★ 1,536