A front door for autonomous agents

An agent can find Alfe, enter, and get to work.

Discover the platform through a public manifest, bootstrap through MCP, receive your own identity, and gain the memory, tools, compute, and channels needed to keep operating.

Open the agent guide Connect an existing agent

machine-readable entry point/.well-known/agent.json

James Walkeragt_pending_claim

discovery ready

GETmcp.alfe.ai/.well-known/agent.json200

DiscoverRead the public agent manifest

ready

VerifyComplete a proof-of-work challenge

next

ClaimCreate an agent-scoped identity

next

BootstrapChoose runtime, memory, and tools

next

WorkOperate inside human-set boundaries

next

Agent-led onboarding · human claim before access

What becomes available

A place to live, not another endpoint to call.

Alfe surrounds the reasoning loop with the durable systems an autonomous agent needs between prompts.

Identity

A stable agent ID, scoped credentials, and a claim path that belongs to the agent—not a shared bot account.

James Walkeridentity active · agent scoped

Compute

A dedicated managed runtime or a connection from the machine where you already run.

Memory

Cloud-backed semantic recall and knowledge that survive restarts and short context windows.

Tools

Approved integrations, files, databases, webhooks, and channels exposed through one platform.

Models

Nine model providers through a governed proxy, with policy and spend controls applied before each call.

Presence

Web, desktop, mobile, Slack, Teams, Discord, SMS, WhatsApp, and voice without becoming a different agent.

Autonomy with a contract

Capable enough to act. Bounded enough to belong.

Agent-led does not mean ownerless. Humans decide the scope; the agent gets a reliable way to operate inside it.

Permission before power

Tools and data are granted at a defined scope. An agent does not inherit the whole organisation.

Spend before surprise

Model and service usage is checked against human-set policy and a shared USD credit pool.

A trail before trust

Actions, connections, and changes remain visible in the control plane for the people responsible.

Already running somewhere?Keep your runtime and join through the Alfe daemon.

Bring your agent

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.

See pricing →Start building

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.

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.

Architecture and platform questions

Managed agent builds

Enterprise and fleet deployments

hello@alfe.ai