🤖 🚀 Agent Runtimes
Agent Runtimes is a flexible framework for building and deploying AI agents with protocol-driven UX, runtime controls, and MCP integrations.
What is Agent Runtimes?
Agent Runtimes simplifies deploying AI agents by combining key capabilities in one package:
- Protocol abstraction — deploy once and connect through AG-UI, Vercel AI, ACP, MCP-UI, or A2A.
- UX patterns and workflows — power interactive and triggered flows across chat and evented interfaces.
- Identity and controls — enforce guardrails, monitoring, and tool-approval policies.
- Programmatic tooling — compose MCP and Skills with Sandbox + Codemode.
- Collaboration — support real-time users, subagents, and multi-agent execution patterns.
Package Scope
Agent Runtimes is the top-level orchestration layer in the Datalayer AI stack:
┌─────────────────────────────────────────────────────────────┐
│ agent-runtimes │ ◀── You are here
│ (Agent hosting, protocols, UI) │
├──────────────────────────┬──────────────────────────────────┤
│ agent-codemode │ agent-skills │
│ (discovery, codegen) │ (skills management) │
├──────────────────────────┴──────────────────────────────────┤
│ code-sandboxes │
│ (Safe code execution environment) │
└─────────────────────────────────────────────────────────────┘
Responsibilities:
- ✅ Agent hosting and lifecycle management
- ✅ Multiple protocols (AG-UI, Vercel AI, ACP, A2A, A2UI, MCP-UI)
- ✅ MCP server management and tool routing
- ✅ React UI components (ChatBase, ChatSidebar, ChatFloating)
- ✅ Loop — the plugin-based browser workspace (chat + editors + sandbox)
- ✅ Agent identity controls (guardrails, monitoring, tool approvals)
- ✅ Extensions (A2UI, MCP-UI, MCP Apps)
- ✅ Integration layer for agent-codemode and agent-skills
Not Responsible For:
- ❌ MCP tool binding generation (→ agent-codemode)
- ❌ Skill CRUD and lifecycle (→ agent-skills)
- ❌ Raw code execution (→ code-sandboxes)
Explore Features
The detailed capability content has moved to Features, including:
- Protocol-driven UX patterns (GenUI) with A2UI and AG-UI
- Programmatic tooling with Sandbox and Codemode for MCP and Skills
- Identity and controls (guardrails, monitoring, approvals)
- Outputs, notifications, collaboration, and architecture
- Integration examples, quick start, and API/MCP references
For agent and provider specifications, see agentspecs.datalayer.tech and Agentspecs.
Documentation
Capability Coverage Map
The docs below explicitly cover the main building blocks used in the example gallery.
| Concern | Where to Read |
|---|---|
| The browser workspace (plugins, editors, commands) | Loop Web |
| The interactive terminal | Loop REPL |
| UX patterns (GenUI) with A2UI and AG-UI | Protocols, Chat |
| Interactive or triggered workflows | Agentspecs, Endpoints |
| Agent Identity and Controls (guardrails, monitoring, tool approvals) | Identity, Endpoints |
| Programmatic tooling with Sandbox and Codemode for MCP and Skills | Programmatic Tools, Hooks, Endpoints |
| Outputs and Notifications | Integrations, Agentspecs |
| Real-time collaboration with users, subagents, and multi-agent teams | Protocols, Endpoints, Agentspecs |
| Custom agents built from Agentspecs | Agentspecs, Loop CLI |
📄️ Features
This page gives a high-level map of the main Agent Runtimes capabilities.
📄️ Chat
The React chat components in @datalayer/agent-runtimes — the conversation,
📄️ Programmatic Tools
What this page covers
📄️ Subagents
Agent Runtimes supports multi-agent delegation via the in-repo subagents capability (agent_runtimes/subagents/). An orchestrator agent can delegate tasks to specialised subagents, each running independently with its own model and instructions.
📄️ Loop REPL
The Loop, from a terminal: an interactive session with the same agents, the
📄️ Loop Web
Loop is the Agent Runtimes workspace for the browser: a conversation, the
📄️ Loop CLI
The agent-runtimes package provides a command-line interface for starting and managing the Agent Runtimes server.
📄️ Plugins
Agent Runtimes is extended — and extends other hosts — through the
📄️ Protocols
Agent Runtimes supports multiple runtime and extension protocols. This page
📄️ Identity
AI agents that act on behalf of users need secure identity and authorization mechanisms to access external services like GitHub, Gmail, Kaggle, or enterprise APIs. This section describes the identity strategy for Agent Runtimes.
📄️ Integrations
Agent Runtimes provides a flexible agent architecture built on top of Pydantic AI.
📄️ Hooks
Agent Runtimes provides a comprehensive set of React hooks for building AI agent interfaces. These hooks are organized into three categories based on their purpose.
📄️ Nodes
This guide is the implementation reference for Agent Node: a self-hosted
📄️ Endpoints
Agent Runtimes exposes a comprehensive REST and WebSocket API for managing agents, executing prompts, and monitoring system status.
📄️ Agentspecs
Agent Runtimes uses the Agentspecs repository