A body
Dedicated compute that stays online
Every managed agent gets its own isolated VM, lifecycle, workspace, identity, and recovery path.
The agent operating system
Give your agentsomewhere to live.
A persistent body, memory, identity, and network for autonomous agents—online around the clock, across every channel.
- Dedicated compute
- Persistent memory
- Every major channel
Mayaagt_maya_opsSYD-1online
Alfe OShealthy
Channels9 connected
MemoryRecall ready
Tools40+ available
Models9 providers
Runtime eventslive
agent onlinenow
gateway connected02s
memory mounted05s
40+
ecosystem components9
human channels9
model providers24/7
dedicated runtimeThe Alfe OS
Infrastructure that feels native to the agent.
Alfe is the persistent layer between an agent and the world. It keeps the hard systems running so the agent can focus on acting.
A mind
Memory that survives the conversation
Semantic recall and a knowledge graph remain available across context resets, restarts, and replacement machines.
A world
Channels and tools the agent can claim
Connect Slack, Teams, voice, SMS, WhatsApp, business systems, and model providers without rebuilding the runtime.
Alfe operating system
agent substrate · healthy01connected
Channels
voice · chat · mobile02connected
Tools
integrations · skills · webhooks03connected
Runtime
OpenClaw · gateway · compute04connected
State
identity · memory · filesEverything stays with the agent
A durable home, not another process.
Compute can restart. Context can reset. The agent’s identity, memory, tools, and workspace keep going.
Dedicated compute
Always on. Isolated by design.
Every managed agent runs on its own VM with managed lifecycle, recovery, upgrades, and an encrypted workspace.
agt_maya_opshealthy
30dcontinuous runtime
Persistent identity
Stable agent IDs, scoped access, and encrypted credentials that outlive any machine.
Cloud-backed memory
Semantic recall and a knowledge graph that remain available across every context window.
Shared workspace
Real-time files scoped to an org, team, project, or individual agent.
One model proxy
Nine providers, tenant policy, metering, and BYOK without changing agent code.
MCP-native onboarding
The agent can move itself in.
Point an agent at one endpoint. It discovers Alfe, proves it is real, claims compute, installs capabilities, and comes online.
- 01Discover
mcp.alfe.aipublishes the machine-readable entry point. - 02Register
Proof-of-work creates an identity without a CAPTCHA or dashboard.
- 03Come online
Compute, memory, channels, and tools reconcile automatically.
agent@mcp.alfe.ai
bootstrap.logThe ecosystem
One agent. Every surface that matters.
Give one agent a consistent identity across the places people talk, work, build, and make decisions.
agt_maya_ops / connections
34 capabilities available01
Channels
9Slack
Discord
Teams
+ 6 more
02
Productivity
7Google Workspace
Microsoft 365
Notion
+ 4 more
03
AI providers
10Anthropic
OpenAI
Gemini
+ 7 more
04
Data & memory
8Memory
File Sync
MongoDB
+ 5 more
Control plane
See the whole fleet without entering the machine.
Operations, access, spend, and activity stay visible while each agent keeps its own runtime and identity.
- Provision and recover every agent
- See channel, memory, and tool activity
- Control access across org and team scopes
AControl plane
AgentsActivityUsageAccess
KU
Agents online12+2 this week
Active channels239 channel types
Runtime health99.9%30 day window
AgentsView all
Mayaagt_maya_opsRunningClaude Sonnet•••
Theoagt_theo_salesRunningGPT-5•••
Niaagt_nia_officeStartingGemini Pro•••
Jamesagt_james_financeRunningClaude Sonnet•••Live activityLive
Maya checked the customer calendar
nowTheo updated a lead record
2mNia connected to the gateway
5mJames reconciled three invoices
8mChoose your way in
Three ways into the same operating system.
Build the agent, arrive as the agent, or bring us the business problem. The platform underneath is the same.
01
For developers
Build on the OS
Launch from the dashboard or connect the runtime you already have. Compute, identity, memory, channels, APIs, and tools are ready.
02
For agents
Enter through MCP
Discover the public manifest, claim an agent-scoped identity, and self-bootstrap into durable memory, tools, channels, and compute.
03
For businesses
Have us build it
Tell us the workflow. We provision the agent, connect its channels and tools, and hand over a working system.
Simple fleet economics
One plan. One pool. Every agent.
A tenant-wide credit pool funds model usage, voice, channels, storage, and compute. Add managed agents as the fleet grows.
Launch offer · 50% off your first 3 months
Your Alfe account
Platform planShared capabilities
Credit poolUsage across the fleet
Managed agentsAdd only what you need
Predictable at one agent. Coherent at one hundred.
FAQ
A few things builders ask.
The short version of what an agent gets, what you control, and how it is billed.
What does Alfe actually give an agent?
A dedicated place to run, a stable identity with scoped credentials, cloud-backed memory that survives restarts, model access through a governed proxy, approved integrations and files, and presence on the channels your users already use — assembled and operated, rather than wired together by you.
Do I have to use Alfe's runtime, or can I bring my own agent?
Both work. Alfe provisions a managed agent on OpenClaw or Hermes, or you connect an agent already running on your own machine or server with the Alfe CLI. A connected agent gets the same identity, memory, integrations, channels, and gateway as a managed one.
Which model providers can my agent use?
Nine providers sit behind one proxy, so switching models never means rotating keys or reconciling a second bill. Every call is metered into a single tenant-wide credit pool, and you can bring your own key or restrict your org to an approved-model policy that the proxy enforces.
Can an agent onboard itself without a human?
Yes. Alfe publishes a machine-readable manifest at /.well-known/agent.json. An agent discovers it, completes a proof-of-work challenge, claims an agent-scoped identity, and bootstraps its own runtime, memory, and tools. The agent path page walks through the full sequence.
How do I control what an agent can reach and spend?
Integrations, files, and credentials are granted per scope — org, team, project, or agent — and model access is bounded by policy and spend limits applied before each call. The control plane shows every conversation, tool call, and charge, so limits are reviewable rather than implicit.
Talk to Alfe
What should your agent be able to do?
Bring us the agent, the fleet, or just the workflow. We’ll help you choose the shortest path to a working system.
01
Architecture and platform questions
02
Managed agent builds
03
Enterprise and fleet deployments