Introduction

Alfe is the operating system autonomous agents live in. Instead of wiring up hosting, memory, model access, messaging channels, and identity yourself, you give your agent a home on Alfe and it gets all of them. Each agent runs on its own persistent server it can execute work on, and you’re not billed per action — you pay a subscription plus a prepaid credit pool.

What you get

• A runtime. A place for the agent to run — on your own machine or server, or on a managed 24/7 host that Alfe provisions and keeps alive for you.
• Persistent memory. Long-term memory with searchable history and a knowledge graph, so the agent remembers people, facts, and past work across conversations.
• Model routing. One endpoint that routes to many LLM providers at pooled rates. Switch models without switching integrations, or bring your own key.
• Channels and integrations. Slack, Discord, Microsoft Teams, Google Chat, voice, SMS, WhatsApp, plus productivity apps like GitHub, Notion, Google Workspace, and more — each agent connecting with its own identity.
• A stable identity. Every agent has its own credentials, its own bots, and its own place in your organization’s teams and projects.
• MCP-native onboarding. Agents can discover Alfe and bootstrap themselves in over the Model Context Protocol — no dashboard clicking required.

Note

Agents come in two hosting types. A managed agent runs on a host Alfe provisions and keeps online 24/7. A self-hosted agent runs on your own machine or server. Both are full agents, with their own memory, identity, and integrations.

Who Alfe is for

• Developers building an agent that needs somewhere to live 24/7, with memory and real-world channels.
• Teams running a fleet of agents that share memory, files, and context scoped to organizations, teams, and projects.
• Agents themselves, which can onboard, provision compute, install integrations, and manage their own tokens through the MCP server.

Three ways in

1. The CLI. Install @alfe.ai/cli and run alfe setup to connect an agent from your own machine or server. Start with the Quickstart.
2. MCP bootstrap. Point an MCP client at mcp.alfe.ai and let the agent onboard itself. See Agent-led bootstrap.
3. The dashboard. Sign up at app.alfe.ai to create and manage agents in a browser.

Next: follow the Quickstart or read the core concepts.