FAQ

Is lockhttp ready to use?

It is a pre-release: v0.1.0 is out, built on lockkit v0.2.1. Try it, report what breaks; flags may still move before a stable release.

Do I need lea?

Yes, for now. lockhttp relies on lea’s topology: each agent container on its own internal network, a proxy container as the only way out, and a source map naming the container behind each address. The gate is a plain HTTP(S) proxy, but the guarantee comes from that topology.

Why not a simple domain allowlist?

Because an allowed domain is allowed for reads and writes. lockhttp keeps reads of trusted hosts free and stops writes, uploads and unknown hosts for a human, with a trace of who asked.

What happens when nobody watches the console?

Requests that need a human are held 60 s, then refused with E_PENDING; the item stays on the console and the same request goes through once you approve it. With no console at all, they are refused at once with E_NO_CONSOLE, while trusted reads continue.

Does lockhttp decrypt my traffic?

Only if you turn it on. Without it, lockhttp decides HTTPS on the host name and port (CONNECT and SNI). With a CA generated per installation (lockhttp ca init), trusted only inside lea’s containers, and the signer that holds its key, the gate decrypts and decides each request. Hosts that pin their certificates go in a splice list and stay decided at the domain.

Can the agent approve its own request?

No. Approvals happen only in the console, on the host, in another account and session. lea’s deck shows a counter of pending requests, never an approve button.

Can I let some requests through without asking?

Yes: a free rule, or an auto rule that auto mode may answer. Both are written and confirmed by you, scoped to a container or to all, and some classes (uploads, messaging, publishing, pushes) can never be auto.

Is there an MCP server?

No, not in v0.1.0. Agents meet lockhttp as a proxy; its refusal pages tell them what to ask of you.

Which licence?

Apache 2.0, like the rest of the Lock family.