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.
-
Passive loop detection MCP server for AI coding agents
MIT -
unloop-mcpgithub.com/protonese3/unloop-mcpupdated this year
MCP server that detects AI fix loops and forces course correction
npm MIT -
agent-guard-mcpgithub.com/mdfifty50-boop/agent-guard-mcpupdated this year
MCP server that detects and prevents infinite agent loops — circuit breakers, pattern detection, and stuck-agent recovery for agentic systems
npm -
unloop-mcpgithub.com/synthet1cc/unloop-mcpupdated this year
MCP server that detects AI fix loops and forces course correction. Error fingerprinting + fix similarity analysis + escalating intervention.
MIT ★ 4 -
Stop AI fix loops with an MCP server that detects stuck coding assistants and forces a new approach
-
Human-in-the-loop LLM Failure Detection & Benchmark Toolkit — Auto-flag LLM failures (hallucination, refusal, factual error) with LLM-as-Judge explanations. Confirm them in a review UI or via MCP agent — then export a citable benchmark validated against expert labels. Compatible with Ollama and Amazon Bedrock models.
-
MCP-native planning & memory system for Claude Code. Features real-time plan→code→document→reflect loops, conflict detection, and enterprise hardening.
npm MIT -
3D-Print-Geniegithub.com/welshDog/3D-Print-Genieupdated this year
Self-hosted AI print guardian for the Anycubic Kobra X — PrintGuard camera detection + Discord alerts + Supabase log + MCP + closed-loop pause spike.
MIT ★ 1 -
Agentic accounts-payable automation for German e-invoicing (EN 16931 / XRechnung): LangGraph agent with human-in-the-loop approval, fraud detection, 3-way matching, MCP server and CI-gated evals.
MIT -
Product-Led Growth (PLG) analysis toolkit that detects tech stacks, plans growth loops and builds the loop iteratively.
npm MIT ★ 129 -
Local-first guard layer for AI coding agents: rejection memory, loop and drift detection, assumption ledger, and evidence-based completion checks. MCP server + CLI.
MIT -
malrang_auto_ctfgithub.com/malrang-me/malrang_auto_ctfupdated this year
Autonomous CTF solving framework for Claude Code — 8 categories, 11 MCP servers, auto loop detection, parallel solving
MIT ★ 1 -
vlkgithub.com/aranajhonny/vlkupdated this year
Temporal state machine for AI agents over MCP: PRESENT/PAST/FUTURE memory, error-loop detection, preventive constraints. Single Rust binary, SQLite.
mcpb ★ 3 -
skill-curator-mcpgithub.com/filhocf/skill-curator-mcpupdated this year
Skill lifecycle intelligence MCP — semantic matching, feedback loop, gap detection, scout
pypi Apache-2.0 -
A line check for hospitality operators' Xero books. Deterministic anomaly detection over 36 months of live data — duplicate payments, cost drift, wage-ratio creep, customer churn — with a Claude layer for plain-English narrative, drafted chase emails, and a live agent over the Xero API + Remote MCP server. No LLM in the detection loop.
MIT -
MCP server giải CAPTCHA 100% local: browser widget + AI image-grid loop + OCR, kèm stealth browser và anti-bot detection
MIT -
mcp-knowledge-basegithub.com/zavora-ai/mcp-knowledge-baseupdated this year
Knowledge Base MCP Server — articles, policies, known issues, TF-IDF search, feedback loops, gap detection, and source-grounded answers
cargo Apache-2.0 -
MCP server connecting Claude to Dexcom CGM data: trends, event logging, AGP charts, parameter-drift detection. Human-in-the-loop by design. Runs on Fly.io.
MIT -
smalltalk-cligithub.com/apexautomate/smalltalkupdated this year
Institutional memory for AI agents. Structured decisions, contradiction detection, per-response drift prevention, and a compounding PAG loop. 25 MCP tools. Zero infrastructure.
pypi MIT -
Spindle: two-deck DJ app with sync, loops, EQ, key detection, key lock and key sync, drivable by an AI agent over MCP. Freeware for macOS, Linux, Windows.
NOASSERTION -
duvo-assessmentgithub.com/obitton/duvo-assessmentupdated this year
MCP server that lets an agent do a category buyer's job: detect stock shortfalls across stores, explain the reasoning, and raise replenishment orders with a human in the loop
-
Model Context Protocol (MCP) server that turns MongoDB operational data into governance actions. Exposes tools for drift detection, compliance checks, bottleneck analysis, and closed-loop remediation via Jira and code agents. No UI, model-agnostic, production-scalable
MIT -
Pre-execution governance for AI agents. Sub-millisecond tool call validation, drift detection, circuit breakers, human-in-the-loop holds. 829 tests. 56 MCP tools.
go MIT ★ 1 -
contextpulse-mcpgithub.com/DIYA73/contextpulse-mcpupdated this year
Real-time MCP server that tracks context budget, detects agent loops, and alerts before token overflow — for Claude Code, Cursor, and any MCP-compatible AI tool.
★ 1 -
AI SRE agent that detects, diagnoses and safely fixes incidents in a microservices app: Kubernetes, Prometheus/Grafana, MCP tools and human-in-the-loop approvals.
MIT ★ 1 -
The-Autonomous-Merchant-Growth-Agentgithub.com/ben854719/The-Autonomous-Merchant-Growth-Agentactive
An agentic AI that improves merchant performance through an Observe, Diagnose, Decide, Act loop, analyzing key signals to compute KPIs, detect bottlenecks, and generate high‑impact actions like pricing updates, product page fixes, discounts, and inventory alerts.
Apache-2.0 -
claudewatchgithub.com/blackwell-systems/claudewatchupdated this year
AgentOps for Claude Code. Monitors AI agent behavior during development: real-time error loops, drift detection, friction analysis. Claude gets self-awareness hooks + MCP tools. You get cost-per-commit, agent success rates, exportable metrics. Like DevOps, but for AI agents.
MIT ★ 9 -
agent-control-planegithub.com/sujithnsn953/agent-control-planeupdated this year
The systems layer around an agent, not another agent. Dual-era MCP server (2026-07-28 + legacy handshake), scoped tool grants with seven detectable failure modes, OTel GenAI tracing, governed memory, hybrid retrieval measured with recall@k, human-in-the-loop approval. Zero runtime deps.
MIT -
memoryvault-kitgithub.com/ayushmall/memoryvault-kitupdated this year
A personal memory layer for your AI tools. Intelligent authoring (gap detection, self-enriching graph, session-synthesis loop) + measurably good retrieval (94.9% blind Cov@10, sub-ms latency). Domain-agnostic framework — works for any context you want to track. MCP-native, Claude-Code-first, works with Cursor/Continue/Cline/OpenAI/Gemini.
MIT ★ 2 -
mcp-dual-cycle-reasonergithub.com/cyqlelabs/mcp-dual-cycle-reasonerupdated this year
A MCP server implementing the Dual-Cycle Metacognitive Reasoning Framework for autonomous agents. A loop prevention and experience recall mechanism.
npm MIT ★ 9 -
Zero-dependency runtime control for production Python agents - stop loops, retry storms, and budget burn
pypi MIT -
Human-in-the-loop browser sessions for MCP clients using Playwright and Chrome DevTools Protocol.
★ 1 -
Vigila modelos de IA en producción y cómo los usan las personas. Se maneja con un asistente por MCP.
MIT -
mcp-dual-cycle-reasonerstale
MCP server implementing dual-cycle metacognitive reasoning framework for autonomous agents
npm MIT -
keeps the rules, decisions, and proof-of-work an agent needs in front of it
npm MIT ★ 2 -
court-notice-gatewaygithub.com/yashpatil582/court-notice-gatewayupdated this year
Ingestion pipeline for US bankruptcy court notices: deterministic phishing gate, LLM extraction with confidence routing, review UI, MCP server
MIT -
A visual inspection agent that remembers: it compares every capture against the stored baseline of the same asset, decides in code what to do next, and stops for a human when a finding is severe. OpenCV 5 (ALIKED + LightGlue), six MCP tools, one arm64 Lambda.
MIT -
Find, explain, and safely fix bad data. Local-first data quality for files, databases, CI pipelines, and AI agents.
pypi Apache-2.0 ★ 2 -
Open-source privacy and approval gateway between AI assistants and your business systems (MCP): human approval, local PII checks and audit for Gmail, Drive, Slack, Salesforce, Jira and more.
pypi Apache-2.0 ★ 2 -
Agentic fraud investigation for Transakt — Kafka + ML scoring, LangGraph over MCP, human-in-the-loop, A2A, Grafana
-
mcp-servergithub.com/agenttrust/mcp-serverupdated this year
MCP server for AgentTrust — A2A communication, agent identity, HITL escalation, and prompt injection detection
npm MIT ★ 1 -
Isse Automate Kardo finds the repetitive workflows hiding in your activity data, shows where the time goes and why, and proposes automating the low-judgment steps. Nothing runs until you say yes. Hexi, its assistant, answers by voice or text through an MCP server.
MIT -
DriftBellgithub.com/LikithSh3tty/DriftBellupdated this year
An ML drift watchman that investigates before it acts and asks you first. n8n owns triggers, credentials and the Telegram approval; a LangGraph agent gathers evidence, critiques itself, and resumes from disk hours later by thread_id. Champion/challenger retraining, RAG chatbot, MCP server. LangGraph, n8n, FastAPI, SQLite, scikit-learn, Gemini.
MIT -
ThumbGate Pre-Action Checks self-improve from ranked lessons and repeated failures, hard-block detected secret leaks, and block matches in strict mode.
npm MIT ★ 31 -
Real-time agentic payment fraud investigation: Kafka + Isolation Forest + LangGraph multi-agent with MCP tools, RAG, human-in-the-loop and eval-gated CI
-
Claude agent that finds, fixes and verifies faults in a simulated 100-vehicle fleet through a Go MCP server. Code-enforced write and stale-data checks, a ground-truth eval harness, and a rule-based baseline: Claude 6/6 vs rules 0/6 when the fix is only in the manual.
Apache-2.0 -
Sentinel-AIOpsgithub.com/Anbu-00001/Sentinel-AIOpsupdated this year
A production-grade AIOps framework focused on model integrity and autonomous reliability. Features a LightGBM-driven multiclass inference engine via FastMCP, validated against data leakage through NMI analysis. Includes real-time Population Stability Index (PSI) drift monitoring and a closed-loop human-in-the-loop feedback system for sustainable ML
MIT ★ 2 -
An enterprise-grade, autonomous AI agent built to detect, audit, and remediate cloud security misconfigurations (like public S3 buckets) in a completely isolated, simulated AWS environment.
MIT -
Pause and safely resume Browser Use tasks at CAPTCHA: same-session detection, authorized handoff/test actions, and fresh validation. Local MCP + SDK.
npm MIT ★ 1 -
kontrol-freek-mcpgithub.com/berkbayri/kontrol-freek-mcpupdated this year
AI assumption firewall for MCP agents — intercepts risky decisions, checks past decisions, detects contradictions, and gates human approval via Telegram or Slack.
pypi ★ 1 -
Foremangithub.com/rohitkumarmanne-442/Foremanupdated this year
The review inbox for your AI workforce - risk-ranked review cards for every agent session (Claude Code, Cursor, any IDE), claims-vs-evidence verification, and ed25519-signed MCP receipts with rug-pull detection
MIT ★ 6 -
Local MCP tools for document search, reusable document profiles, spreadsheet SQL, JSON snapshot changes, resumable API batches and retry-loop checks. Built by AVRG3.
mcpb MIT -
Precision-first C/C++ static analyzer on Clang LibTooling — catches the null-derefs, leaks and overflows your compiler waves through, each with a dataflow trace. MCP-ready for the AI coding loop.
Apache-2.0 ★ 2 -
AI-assisted hospital insurance claim review. LangGraph multi-agent pipeline (extraction, policy RAG validation, fraud scoring, summary) with human-in-the-loop decisions. FastAPI + Next.js 16 BFF, ChromaDB, MCP server, 3D UI.
-
Flight recorder, MAST failure attribution & on-call ops loop for AI agents: replay, tamper-evident evidence, triage queue, letter grades, handoff briefs, fleet tail, price catalog, conversation recall. 53 MCP tools. Zero dependencies, Python 3.10+.
MIT ★ 1 -
Validated analytics agent for CSV/Excel. LLMs plan - as a JSON plan or a bounded tool-calling loop - while deterministic pandas/scipy/DuckDB computes and re-verifies every number. No arbitrary code execution: the model only selects audited operations.
MIT ★ 4 -
Persistent memory + workflow orchestration as an MCP server. 78 canonical tools across 21 namespaces, SQLite-backed, with an LLM-as-judge drift gate and a vibe-flow loop (spec to artifact to drift to publish).
npm NOASSERTION ★ 2 -
Deciding fault takes a record outside the agent's failure domain. ThetaCog writes one: a receipt showing whether the instruction or the work left the lane, read from the commit against the spec declared before the run, with zero models in the loop. Free to run (MIT); patent pending (US Patent App. 19/637,714).
npm MIT ★ 1