#sandbox

1 place to call all your agents - OpenCode, Hermes, Claude Managed Agents, Cursor Agents API, DeepAgents.

Fast cold-boot MicroVM sandboxes for AI agents on cocoon

Run multiple agents in parallel sandboxed VMs, with a single command, on your PC or in the cloud

Sandboxed devcontainer for running Claude Code, Codex, and OpenCode in bypass / auto-approve mode.

The runtime substrate for agentic systems — one Go binary, six LLM providers, MCP-native, configurable as a managed sandbox or full agentic dev environment. Where agents live, talk, and learn.

Your dashboard is data. Any AI can build it; any human can edit it. A protocol + runtime for agent-composable dashboards — layout-as-data, one guarded control plane, sandboxed agent-authored widgets, MCP server included.

A quantum prompt playground & macOS-native Rust coding agent: morphing fluid entropy into rigid execution checkpoints. Entropy cannot lie.

An open, community- and AI-agent-driven Minecraft-style voxel engine in Rust on wgpu.

Skainet Bridge — local automation bridge for AI agents: one process, one port. MCP server + client, REST, browser extension, sandboxed exec. Independent project, unaffiliated.

Run an AI second brain and your repositories on a Linux VM you control. One Go binary: no daemon, no credentials, and a backend that cannot push.

Firecracker-microVM sandbox for Claude Code: ephemeral hardware-isolated runs (~0.4s boot) + content-addressed environment stages, as MCP tools and a CLI

A sandboxed, full-featured agent operating system in a single Go binary. Bring any model, manage local models, point it at a goal, and grant it real authority: every action is sandboxed, governed, and sealed into a verifiable, tamper-evident record an independent party can check. Runs interactive or 24/7, or embed it in your own system.

A general-purpose agent fleet you run yourself — any model, in a sandbox, on a budget, connected to your data.

Autonomous agents with receipts.

Run Claude Code on Windows inside a real Linux sandbox. WSL 2 + ai-jail (bubblewrap, Landlock, seccomp) so the agent can only touch your project, with SSH commit signing still backed by Bitwarden.

AI security layer for the agentic internet — sandbox isolation, Civic governance, and a one-API-call clearance system for any OKX.AI agent.

Talunor is a local-first terminal AI agent written in Go, with persistent, auditable SQLite memory, guarded tool use, and a complete step-by-step course explaining how it is built.

Confidential AI agent runtime — hardware-attested, governed sandboxes (AMD SEV-SNP). Apache-2.0. Rust top-to-bottom.

🧩 Open-source control plane for running AI agents in secure, observable sandboxes.

Durable coding-agent sessions on Cloudflare, with sovereign local libkrun microVM handoffs.

Autonomous coding agent that fixes buggy repos by iterating against their pytest suite in an isolated Docker sandbox provider-agnostic LLM layer, 22-task eval harness, and OpenTelemetry→Langfuse tracing.

Run Claude Code with --dangerously-skip-permissions inside a container with no route to your host and SNI-filtered egress

Run AI-generated HTML tools in an isolated sandbox. Static capability scanning, per-tool origin isolation, dynamic CSP enforcement.

Claude Code skill: safely spawn a sandboxed, phone-drivable Remote Control session

Anchor Browser is an AI-native cloud browser infrastructure platform that lets AI agents interact with the web the same way a human would. It provides hosted, isolated Chromium sessions ("Anchor Chromium"), built-in stealth and bot-evasion tuning, an authentication layer (OmniConnect) for managing logged-in user credentials, a built-in enterprise…

DeepSeek Harness 自动审批插件：在 workspace-write 沙箱之上叠加确定性规则 + LLM 安全审批，自动模式不放宽沙箱、fail-closed。 Safe auto-approval for DeepSeek Harness — deterministic rules + LLM review on top of the workspace-write sandbox. Auto mode without ever granting full-access.

Run Claude Agent SDK agents in sandboxed Cloud Run Jobs inside your own GCP project — audit trails, human approval gates, agents, scheduled deployments, platform connectors, Agent Skills, and zero always-on cost

An AI team for a company of one. You do three things: say what you want, approve the plan, merge the PR. Nine roles do the rest — planning, building, reviewing it twice — each inside its own container, so if an agent runs rm -rf it happens to a container, not to you.

Terminal agent with durable sessions and background jobs, a fail-closed filesystem sandbox, and first-class unattended runs. Written in Go.

Trust-bounded acceptance harness for Claude coding agents — sanitized workspaces, typed tools, isolated oracle evaluation, evidence receipts.

Bubblewrap (bwrap) sandbox for Strands Agents — unprivileged Linux user-namespace command/code execution.