Best AI GitHub Repos

Skills for creating MCP Apps with the MCP Apps SDK

# mcp-tunnels Connect Claude to an MCP server running inside your private network through an Anthropic [**MCP tunnel**](https://platform.claude.com/docs/en/agents-and-tools/mcp-tunnels/overview) — no inbound ports, no public exposure, no IP allowlisting on your origin. Traffic flows over an outbound-only connection. > **Research preview.** MCP tunnels is provided "as-is" with no uptime or > support commitment and depends on a third-party transport provider > (Cloudflare). Review the > [securit

Mercado Pago full-product integration toolkit. One agent routes to four orchestration skills (mp-integrate wizard, mp-webhooks, mp-test-setup, mp-review) that pull every endpoint, payload, and snippet live from the official Mercado Pago MCP server. The MCP must always be connected — there is no offline mode.

Mergify Community Issue Tracker

Secure access to Miro boards. Enables AI to read board context, create diagrams, and generate code with enterprise-grade security.

Keep your coding agent up to date with the latest web best practices

Run your monday CRM in plain language. Build a pipeline from scratch, start the day with a ranked deal briefing, spin up a forecast dashboard, audit board health, clean up messy data in bulk, and turn meeting notes into deal updates. Every skill writes back into monday as a real update, doc, or dashboard. Built on the official monday MCP connector.

NetSuite agent skills from Oracle. Authoring guidance for SuiteCloud Development Framework (SDF) objects and UIF single-page-app components, plus runtime guidance for the NetSuite AI Service Connector.

New Relic observability intelligence for Claude Code. Investigate APM performance, analyze cloud costs, debug Kubernetes, write NRQL queries, and respond to alerts — all from your terminal using New Relic telemetry data.

NVIDIA agent skills for accelerated-computing workflows — starting with cuOpt vehicle-routing optimization (VRP, TSP, PDP) via the cuOpt Python API.

Drive the Oracle AI Data Platform (AIDP) Databricks Migration Toolkit in natural language. Plans and executes automated Databricks → AIDP migrations of notebooks, jobs, schedules, and catalog DDL — Pass-1 dependency rewrite + Pass-2 cell-by-cell execute/verify/fix on a live AIDP cluster with Claude.

Oracle AI Data Platform (AIDP) Workbench engineer agent for Claude Code — a 37-skill agent that operates the full Spark/Delta lakehouse in natural language. Discovers your catalog into a grounding cache, turns plain English into accurate Spark SQL, and runs the lifecycle (CREATE/INSERT/UPDATE/DELETE/MERGE, OPTIMIZE/VACUUM, time-travel). Ingests files, profiles data and sets quality rules, authors and repairs pipelines, provisions clusters, and debugs via the Spark UI. Governs the platform (roles, credential store, Delta Sharing, audit logs), plus native Git, bundles, and MLOps/MLflow. Runs via the official Oracle aidp CLI.

Oracle AI Data Platform Workbench Spark connectors for Claude Code. 18 connector skills covering every data source workbench customers commonly need: Oracle Autonomous DB family (ALH/ADW/ATP) via wallet/IAM-DB-Token/API-key, ExaCS, Fusion ERP REST, Fusion BICC, EPM Cloud Planning, Essbase 21c, OCI Streaming (Kafka), OCI Object Storage, Apache Iceberg, plus external systems (PostgreSQL, MySQL/HeatWave, SQL Server, Snowflake, Azure ADLS Gen2, AWS S3, generic REST, custom JDBC, Excel). Live-validated on the workbench `tpcds` cluster (Spark 3.5.0): 17 PASS / 4 ship-as-is out of 21 test rows.

Output.ai workflow development toolkit for Claude Code. Adds 5 specialist agents (planner, builder, debugger, prompt writer, quality reviewer), 40+ slash-command skills covering scaffolding, debugging, evaluation, and credential management, plus a SessionStart hook that auto-loads Output SDK conventions so Claude understands the framework before the first prompt.

Analyze business data and build custom Pigment models, metrics, and boards through natural language.

Unified multimodal backend for AI data apps

Preset CLI skills for explicit shell, scripting, and CI/CD workflows driven by the `sup` CLI (PyPI package `superset-sup`). Use only for CLI workflows; do not use for MCP-only work or as a substitute for direct API calls when the user wants HTTP/SDK code.

# project-artifact Generate and publish a **living status page** for a project that's too big for one update — a migration, a launch, a research effort, anything with several workstreams tracked over time. The page is a single self-contained tabbed HTML file (overview & success criteria, the workstream sequence, an always-visible "Next steps" strip, plus background / plan / risks / FAQ tabs when they earn their place), published with Claude Code's built-in `Artifact` tool to a private `claude.a

AI Agent Framework, the Pydantic way

Agent skills for Qdrant vector search covering scaling, performance optimization, search quality, monitoring, deployment, model migration, version upgrades, and SDK usage across Python, TypeScript, Rust, Go, .NET, and Java.

Qodo Skills provides a curated library of reusable AI agent capabilities that extend Claude's functionality for software development workflows. Each skill is designed to integrate seamlessly into your development process, enabling tasks like code quality checks, automated testing, security scanning, and compliance validation. Skills operate across your entire SDLC—from IDE to CI/CD—ensuring consistent standards and catching issues early.

Agentic engineering skills for Qt software development — Qt C++/QML code review, QML coding, and Qt C++/QML code documentation.

MCP server for AI coding agents to create, manage, and interact with Quarkus applications. Provides tools for project scaffolding, dev mode lifecycle, extension skills, Dev MCP proxy, and documentation search.

# Receipts Generate a personal Claude Code impact report — "receipts" — from your own session transcripts, for the conversation where someone asks what all this Claude Code usage is actually buying. <img src="assets/sample-receipt.png" width="400" alt="A printed-receipt-styled report headed 'Claude Code — usage receipt, Morgan Lunt', covering 2026-06-21 to 2026-07-20, active on 28 of 30 days. It lists 40 sessions, 64 prompts, 14 files touched, 3,120 lines touched, 12 commits carrying that wo