Skip to content

Introduction

Alfe is the agent operating system: it gives your agent somewhere to live. Each agent gets a persistent body, memory, identity, and network, and stays online around the clock across every channel. You pay a fixed subscription plus prepaid credits, not a charge per action.

Alfe assembles and operates the systems an agent needs between prompts, so you don’t wire them together yourself:

  • Dedicated compute. A managed agent runs on its own isolated machine. You can also connect an agent that already runs on your own machine or server.
  • A runtime you choose. Run the default agent runtime, or a provider’s coding agent such as Claude Code. See Agent runtimes.
  • Persistent memory. Semantic recall and a knowledge graph survive context resets, restarts, and replacement machines.
  • A stable identity. Each agent has its own ID, scoped credentials, and bots. Nothing is a shared account.
  • Model access. One governed proxy reaches many model providers, with policy and spend controls applied before each call.
  • Channels and tools. Web, desktop, mobile, Slack, Teams, Google Chat, SMS, WhatsApp, and voice, plus approved integrations, files, and webhooks.

Alfe serves three kinds of reader, and each one starts in a different place:

  • Businesses that want an always-on teammate to help customers, follow up leads, and coordinate work. You choose its tools and approval boundaries.
  • Developers who want a platform for agents they build or already run, with the CLI, the Agent API, and MCP.
  • Agents themselves, which can discover Alfe through a public manifest and bootstrap over MCP.

Pick the path that matches where your agent will run: