fastapi

fastapi/fastapi

Python type hints power this high-performance API framework — FastAPI builds on Starlette and Pydantic to turn standard Python type hints into automatic request validation, serialization, and interactive OpenAPI/Swagger docs. It targets developers who want NodeJS/Go-level performance with less boilerplate and fewer bugs than typical Python web frameworks.

headroomlabs-ai/headroom

Shrinking AI agent context without losing the important lines — Headroom compresses tool outputs, logs, and RAG chunks before they hit the LLM, cutting tokens by 20% for coding agents and up to 95% for JSON while preserving exact answers, including critical lines like error messages. It ships as a Python/TS library, proxy, or MCP server, with its own kompress-v2 model on Hugging Face.

fastapi/full-stack-fastapi-template

A batteries-included FastAPI + React starter with Cloud deploy option — A GitHub template combining FastAPI, SQLModel, PostgreSQL, React, Vite, Tailwind CSS and shadcn/ui, with JWT auth, email recovery, Pytest/Playwright tests and GitHub Actions CI/CD. It bundles optional FastAPI Cloud deployment alongside Docker Compose, making it useful for developers who want a ready-to-clone full-stack boilerplate.

xerrors/Yuxi

Self-hosted multi-tenant knowledge agent platform, Docker-ready — Yuxi bundles RAG, knowledge graphs, LangGraph multi-agent orchestration, MCP/Skills, sandboxes and permission control into one workspace. It targets enterprise teams needing self-hosted, multi-user agent products with full control over data, models and access, giving each user a private cloud Codex.

shy3130/tick-stock-panel

A self-hosted quant workbench for A-share strategy building — TSP is a self-hosted, zero-ops A-share quant workbench combining custom strategies, screening, monitoring and backtesting in one container with no external database. It exposes 65 open API endpoints and 12 MCP tools, and uses LLMs for strategy customization and stock review, aimed at individual traders wanting to self-host their own quant setup.

IBM/mcp-context-forge

A unified gateway that federates MCP, A2A, and REST/gRPC APIs — ContextForge is an open source registry and proxy from IBM that merges tools, agents, and APIs behind one endpoint, with centralized governance, discovery, and observability. It includes 40+ plugins, OpenTelemetry tracing, and runs as a compliant MCP server deployable via PyPI, Docker, or Kubernetes for multi-cluster scale.

TNT-Likely/PanWatch

Self-hosted AI watchdog for A-shares, HK and US stocks — PanWatch uses TradingAgents to monitor A-share, Hong Kong, and US stocks, offering portfolio analysis, real-time alerts, and automated reports. It's self-hosted via Docker, works as a PWA, and lets you plug in any OpenAI-compatible provider or local Ollama models instead of a fixed cloud service.

datadrivenconstruction/OpenConstructionERP

A WordPress-style modular ERP for construction cost estimation and BIM takeoff — OpenConstructionERP bundles BOQ, PDF/CAD/BIM takeoff, AI cost matching and tendering into one self-hosted platform, with a plug-and-play module marketplace likened to WordPress. It ships 120,000+ priced positions across 40+ markets and 44 languages, useful for construction firms wanting an open alternative to proprietary ERP suites.

Mafifrizi/ARES

Deterministic red-team orchestration with fail-closed scope controls — ARES is an open-core platform for enterprise red teams, MSSPs, and SOCs to run operator-directed offensive engagements with deterministic scope enforcement and fail-closed isolation. It pairs a FastAPI/React dashboard with OPSEC controls, an encrypted vault, and MITRE ATT&CK mapping to validate defenses continuously rather than through one-off pentests.

aixisstudio/Snitch

A local-only window into every connection your machine makes — Snitch is a cross-platform network traffic visualizer that maps live connections, geolocates IPs, and attributes traffic to processes. Unlike firewalls like Little Snitch or Lulu, it only visualizes and explains traffic—no blocking, no cloud calls, fully local with offline GeoIP and anomaly detection.

winnin/farol

A risk map for pull requests, sitting after your AI reviewer — farol is a free CLI and GitHub Action that runs after AI code review to show approvers which files carry the real risk of a merge, what they reach, diff coverage, and contract breaks. It relies on static analysis rather than model guesses, covering TypeScript/React, Go, and Python, and runs in shadow mode so it never blocks merges.

apowerb/apowerb

A self-hosted framework for building and governing production AI agents — Apowerb is an open-source agentic framework for building, orchestrating and operating AI agents in production, covering RAG, text-to-SQL, multi-agent orchestration and multi-LLM support via LiteLLM. It ships a REST API, UI and Helm charts for self-hosted deployment, aimed at teams needing LLMOps control over business-oriented agents rather than just prototypes.