ollama/ollama
Ollama is a versatile platform that enables easy deployment, customization, and management of large language models locally across multiple operating systems with support for various model formats and a REST API.
Awesome AI Agents › AI Agent Deployment
Th0rgal/sandboxed.sh
sandboxed.sh, formerly Open Agent, is a self-hosted backend written in Rust that executes autonomous coding agents inside isolated Linux workspaces built on systemd-nspawn or Docker. It is deliberately only half of a system. A coordinator decides what to do and when, and drives sandboxed.sh over MCP; the reference coordinator is the authors fork of Hermes with a Python gateway, a CLI and an Electron desktop app, but any MCP-capable assistant can take that role. The coordinating agent never runs untrusted code itself: each unit of work is handed over as a mission, executed in a throwaway workspace under a chosen harness such as Claude Code, OpenCode, Codex, Gemini or Grok, and streamed back as structured results. Four concepts organise the system. A project is the durable unit of work, holding a mode, an autonomy grant covering merge authority, budget and parallelism, plus tracks and open decisions, stored in a local database and served over an HTTP API. A controller is a scheduled coordinator job that wakes up, reads its control conversation, GitHub state and project state, dispatches missions and reports structured status trailers; controllers can launch missions on another project they depend on. A conversation is the durable chat thread bound to a project. A mission is one isolated agent run that writes code, runs builds and opens pull requests, tagged with its project and track. Controllers write project state through MCP tools rather than free text, and a state ingestor folds status trailers back into the project record. The same roster is rendered by a web dashboard, the desktop projects board and an iOS app. Skills, configuration and encrypted secrets are kept in a git repository. Docker and native installation guides are provided. It targets people running long, unattended agent operations on their own hardware.
https://github.com/Th0rgal/sandboxed.sh
Also in AI Agent Deployment
alibaba/nacos
Nacos is a dynamic service discovery, configuration, and service management platform designed to simplify building and managing cloud-native applications and microservices.
e2b-dev/E2B
E2B is an open-source secure cloud runtime infrastructure that enables safe execution of AI-generated code in isolated sandboxes for AI applications and agents.
GoogleCloudPlatform/agent-starter-pack
A collection of production-ready Generative AI Agent templates for Google Cloud that accelerates development by addressing deployment, evaluation, customization, and observability challenges.
zebbern/claude-code-guide
Claude Code Guide is a comprehensive and authoritative resource providing detailed instructions, tips, and references for using the Claude Code CLI tool to interact with the Claude AI model, including advanced features and MCP integration.
boxlite-ai/boxlite
Rust engine giving AI agents hardware-isolated persistent micro-VMs that run any OCI image, embeddable as a daemonless library in Python, Node, Go, Rust or C, or run as a CLI, REST server or cloud control plane.
jim-schwoebel/awesome_ai_agents
A comprehensive and community-driven repository offering over 1,500 curated resources, tools, and projects related to AI agents for various applications and development needs.
Forethought-Technologies/AutoChain
AutoChain is a lightweight and extensible framework that simplifies building and automatically evaluating custom generative agents powered by large language models using custom tools and simulated multi-turn conversations.