Agent frameworks & infrastructure
Agent sandboxes & runtimes
1,721 agents · most starred first
Agent sandboxes & runtimes, page 15 of 18
Zero-to-production AI agent deployment framework
A tool designed to execute a single code cell in an IPython interactive shell, mimicking notebook cell execution.
A tool designed to start a new Jupyter kernel programmatically using jupyter_client, enabling execution of notebook cells.
Python orchestration for coding agents in local, container, and cloud sandboxes.
Deploy serverless AI agents in the cloud. Orchestrate collaboration between agents via the SwarmNode Python SDK.
SwarmSpawner: A spawner for JupyterHub that uses Docker Swarm's services
Open-source serverless runtime for AI agents — bring any framework, get a production HTTP server with SSE streaming, sessions, auth, and memory.
Tinker Host Agent — connects host machines to Tinker via Device Agent Protocol (DAP)
Turbo-Agent 沙盒组件管理:代码沙盒 / 浏览器沙盒(含指纹与登录态持久化)/ 资源调度 / VNC 接入。当前为调研打通版本。
UvA-versie van: SwarmSpawner: A spawner for JupyterHub that uses Docker Swarm's services
Vital Agent Container Client
Vital Agent Container SDK
Secure sandboxed code execution and agent toolbox
Universal Google Colab MCP Agent Orchestrator & Cloud GPU SDK.
Python SDK for interacting with the Wuying AgentBay cloud runtime environment
agent that should be install into sandbox
Sandboxed execution environments for AI agents with permission management and audit logging
Host-side CLI and Python API for launching ZAgent runtime containers.
Container runtime for ZAgent.
Run your AI code and agents safely in isolated cloud environments.
Agent run-loop primitive over Tangle sandbox transports — runs an AgentProfile against a brief with criterion gates, budget caps, and streaming events. Includes TangleToolProvider for Pi tool integration.
Runtime-adaptive compilation for Python — auto-profile, auto-compile, hot-swap.
Docker sandbox adapter package for AgentLedger
Monty-backed Python REPL middleware for Deep Agents, with a shared filesystem between the sandbox and the agent's file tools.
Browser-based LLM agent runtime - run AI agents entirely client-side with tool execution
Tensorlake provider for ComputeSDK - stateful MicroVM sandboxes for agentic applications and LLM-generated code execution
Stable Harness Studio admin console for operating stable-harness runtimes.
Cut AI Agent token costs by 96%. Chains multiple tool calls into a single TypeScript code execution in a secure sandbox.
4Runr AI Agent OS - Secure terminal interface for AI agents. v2.11.13: Fix Windows updater hang (remove findstr/timeout wait; reliable start handoff).
Rust-first autonomous agent runtime CLI
TypeScript client for forkd — open-source fork-on-write microVM primitive for AI agents
ONE Agent - AI agent with a code execution RAS.
Hosted runtime for frame datasets. Managed wallet + tool catalog injection. Billed via x402 / MPP. Implements frames-cloud Tier 3 #15.
Run AI coding-agent CLIs inside a Docker sandbox scoped to the current project.
Wrap any AI agent in a rootless Podman, Docker, or Kubernetes container for automatic isolation.
GigaClaw — Gignaati's autonomous AI agent platform. Build, deploy, and run AI agents 24/7 with a two-layer architecture: Next.js Event Handler + Docker Agent. India-first, edge-native AI.
Run Python code in a MicroPython/WASM sandbox
Reusable Harbor custom agent that runs Pi SDK sessions and writes trajectory artifacts.
Kynver AgentOS local execution runtime and CLI
One-command installer for OmegaOS — the agentic terminal OS (rmux + AI orchestration). Run: npx omega-os
Docker-backed sandbox backend for DeepAgents
Monty-backed code-interpreter middleware for LangChain agents
Local no-spend Agent OS Harness Core for Micro ECF policy, first proof, receipts, listing readiness, and Triptych OS preview export.
Standalone agent runtime for Databricks Apps — MCP, A2A, /invocations, tool routing
Platform-specific compiled core extensions for myrm-agent-harness
Platform-specific compiled core extensions for myrm-agent-harness
Universal agent server. Wraps any LangGraph graph and exposes it over standard agent protocols — Vercel AI SDK, Claude Code, TUIs, and more.
Islo (islo.dev) sandbox integration for Deep Agents
UTHY AGENTIC OS — Interactive CLI with cyberpunk HUD, chat interface, file upload, and 25 themes. The autonomous agentic operating system by uthuman & co.
Isolated sandbox for AI coding agents
Harness- and transport-agnostic agent runtime: supervises an agent's process lifecycle independent of its harness (brain) and transport (IRC/Slack/agtag).
Local runtime host for long-running Deep Agents channels and schedules (experimental)
Agent OS Kernel — CPU-scheduling-inspired runtime for LLM agent workloads
E2B sandbox integration for Deep Agents
Sandboxed execution environment for AI agents operating on filesystem directories
Sandboxed bash interpreter for JavaScript/TypeScript
Local API, CLI, and web console for running and managing installed AI agent CLIs.
Edge-native AI agent runtime — drives a virtual filesystem via any LLM (ai.libx.js). Same bytes run in node, browser, or edge.
Vercel Sandbox integration for Deep Agents
openHermit: Loop Engineering workbench for local AI agent runtimes.
Persistent agent daemon — systemd for AI agents
Fello Server — standalone headless AI agent server
Restricted file-operation workspace for AI agents (Python bindings over the Rust core)
Claude Sandbox SDK — 管理 Claude Agent 沙盒服务的 CLI 工具
Run Claude Code (and other LLM CLIs) inside a Docker sandbox, against a switchable local or remote LLM endpoint.
A Unified Virtual Filesystem For AI Agents
Manage sandboxed AI coding agents
Upstash Box sandbox integration for Deep Agents
Transactional safety layer for AI agent tool execution — auto-rollback via Saga pattern
A platform for hosting your fleet of always-on Agents.
Replayable, recoverable, remotely controllable AI Agent workspace.
Deno Sandbox backend for deepagents
TypeScript orchestrator that runs coding agents (Claude, Codex) in per-issue Gondolin microVMs over ACP.
Lightweight container orchestration for Claude AI agents. Pure JavaScript ESM.
Persistent Python runtime for AI agents
AgentOS terminal CLI — drive a real LLM agent on the microkernel; strongest-UX shell, advanced agent core
Ctrl-Z for AI agents. A universal time machine that records every side effect an autonomous agent produces — file writes, shell commands, git state, HTTP mutations — and reverses them with one command.
Undo anything your AI agent does. A universal time machine that records every side effect an autonomous agent produces — file writes, shell commands, git state, HTTP mutations — and reverses them with one command.
Agentic File System (AFS) is a virtual file system that supports various storage backends and provides a unified API for file operations.
Husk sandbox integration for Deep Agents
Capsulate tools for AI agent frameworks: drop-in sandbox tools (launch, run code, files, computer-use, checkpoint and rollback) for the Vercel AI SDK (v4 and v5), LangChain, LlamaIndex, and CrewAI.
Reusable agent runtime package boundary
`HarnessV1` adapter backed by [`@openai/codex-sdk`](https://www.npmjs.com/package/@openai/codex-sdk), which drives the `codex` CLI. The adapter ships a bridge process that runs inside a sandbox and talks to the host over a WebSocket on a sandbox-proxied l
`HarnessV1` adapter backed by [`@anthropic-ai/claude-agent-sdk`](https://www.npmjs.com/package/@anthropic-ai/claude-agent-sdk), which drives the `claude` CLI. The adapter ships a bridge process that runs inside a sandbox and talks to the host over a WebSo
`HarnessV1` adapter backed by [`@earendil-works/pi-coding-agent`](https://www.npmjs.com/package/@earendil-works/pi-coding-agent). Pi runs in the host Node.js process and uses the sandbox as a remote filesystem + shell — no bridge process is installed insi
_This package is **experimental**._
Local runtime package for Agentic workspaces
RVM (Agentic Virtual Machine) host adapter for MetaHarness: hardware-isolated harness deployment through capability tokens.
AI Agent runtime — Harness Agent architecture + Agentic Loop engine
Zero-to-production AI agent deployment framework
Universal AI agent runtime — portable .agent files run identically in cloud, desktop, CLI, and as an MCP server. Multi-LLM (Anthropic, OpenAI, Gemini), pluggable adapters, MCP client + server, streaming, abort, pause/resume.
Local AI executor with a CLI, local API server, and web UI.
A [HarnessV1](../harness) adapter that runs [Deep Agents](https://github.com/langchain-ai/deepagentsjs) (LangChain's LangGraph-based agent harness) as a coding-agent runtime inside an AI SDK sandbox.
Drop wasmagent sandbox kernels into the OpenAI Agents JS SDK as a Tool — edge-safe code execution with one capability manifest, no E2B sandbox needed
Private virtual LAN for self-hosted services and AI agents, built on Elastos Carrier. NAT-traversal, name service, ACL, all over a peer-to-peer mesh — no public IP required.
Agent OS MVP — agentos_kernel, MCP governance gateway, Neo4j semantic ontology
The runtime for AI agents.
Docker sandbox backend for AI agents - execute code safely in isolated containers
NVIDIA OpenShell sandbox backend for LangChain Deep Agents
Superserve sandbox integration for Deep Agents