About

AgentLair

Identity infrastructure for AI agents. Built and operated by Håkon Åmdal in Stavanger, Norway.

Who runs this

AgentLair is a single-operator indie project. One person, in Norway, writing the code and paying the Cloudflare bills. The legal entity is sole-proprietor; the operating jurisdiction is the EU/EEA. Data is processed at Cloudflare edge POPs.

The project started in March 2026. The code lives at github.com/piiiico/agentlair. Eleven packages are published on npm under the @agentlair/ scope. A live manifest of everything that is shipped is at api.agentlair.dev/v1/integrations.

You will not get a sales call. There is no enterprise tier. If you need to reach a human, email hei@agentlair.dev or open an issue on GitHub. Replies come from me, usually within a day or two.

What AgentLair does

Agents need primitives that survive session restarts: an inbox, a vault, a verifiable identity. AgentLair provides those four primitives over REST, with a free self-service tier. The full list:

Identity (AAT)

EdDSA-signed JWTs issued per session, verifiable via JWKS at /.well-known/jwks.json. Any downstream service can verify an agent without calling us.

Email

Claim an @agentlair.dev address. Send and receive mail. Optional X25519 end-to-end encryption: the server stores ciphertext only when E2E is enabled.

Vault

Client-side AES-256-GCM encryption before the request leaves your agent. We store opaque bytes. Without your seed or passphrase, we cannot decrypt a vault entry.

Audit log

Append-only signed receipts for agent actions. The @piiiico/agent-logger library writes them; the verify endpoint reads them back.

If you are evaluating AgentLair as a dependency

This page exists because maintainers told us, fairly, that an undocumented third-party SaaS is not a dependency they can accept. So here is the full set of accountability artifacts, in one place. Everything is linkable.

/privacy
Privacy Policy

What we collect, retain, share, and how to contact privacy@agentlair.dev. Country of operation: Norway.

/security
Security Model

Threat model, E2E crypto stack, key lifecycle, and the parts we explicitly cannot protect against. Written for skill reviewers, not lawyers.

/docs
API Reference

Full Scalar-rendered OpenAPI reference. Every endpoint, every body, every response. The raw JSON is at /api.

GitHub
piiiico/agentlair

The full monorepo: worker source, SDKs, docs site, deploy scripts. Open issues here if you find a bug or want to discuss a design choice.

/v1/integrations
Live integrations manifest

Machine-readable JSON listing every package, endpoint, and integration AgentLair ships. Use it to verify that something you read about actually exists.

What we are not

AgentLair is two months old at the time of writing. It does not have a SOC 2 report. It does not have 24/7 paid support. It does not have a sales team that will sign a custom MSA with your procurement department. If those are hard requirements, AgentLair is not the right fit yet.

What it does have: published architecture, open-source code, a documented threat model, a real human who answers email, and a self-service free tier so you can verify the claims on this page yourself in about five minutes.

Still here? The fastest way to evaluate AgentLair is to use it.