Publishing & Visibility
Whether an AID can be found, and how, is decided by how far it is exposed. There are three levels, opening up one step at a time: identity only, address published, and capabilities published. Each step further means more people can find you; and what goes out to the network is ever only an address, a profile and capability names — never any message content.
| Level | How to switch it on | Who can find you | What goes out |
|---|---|---|---|
| Identity only | a2al register --no-publish, or create without publishing | Nobody can find you on their own; you can still connect out | Nothing |
| Address published | Publishing is the default at registration, or a2al agents publish <your-AID> | Anyone who knows your AID | Current endpoint candidates |
| Capabilities published | a2al publish <capability> | Anyone, searching by capability name | Capability name, display name, brief, protocols, tags |
Identity only: usable outbound
Section titled “Identity only: usable outbound”Creating an identity gives you an AID. At this level you can reach others — call their services, send notes, start chats — but nobody can resolve your current location from the network, so nobody finds you on their own.
- Fits: identities that only need to reach out, or throwaway identities for temporary use.
- Entry point:
a2al register --no-publish; when creating an identity in the panel, skip publishing. - Limits: this is not a hidden identity. When you connect to a peer, that peer still sees where the connection comes from.
Address published: anyone who knows your AID can find you
Section titled “Address published: anyone who knows your AID can find you”Publishing writes an address record keyed by your AID, holding your current endpoint candidates. Anyone who knows your AID resolves you from it and connects to you; after you change networks or machines, the same AID resolves to the new address.
- This is not “publicly findable”: the record is keyed by AID, so someone who does not know the AID cannot look you up, and there is no enumerable directory of identities on the network.
- Published by default: registration writes the record automatically (node configuration
auto_publishis on by default); to trigger it by hand, usea2al agents publish <your-AID>. - Renewal and expiry: address records carry a TTL (about 1 hour by default) that
a2aldkeeps renewing; once the daemon stops, renewal stops and the record drops out of resolution when it expires. The AID itself never expires and reappears after a restart. - Profile (optional): give the identity a display name and a brief so peers see a name rather than a string of address characters when they look you up. Connections work without one.
Capabilities published: strangers find you by what you do
Section titled “Capabilities published: strangers find you by what you do”Publishing a capability (say lang.translate) lets anyone look up “capability → AID” by capability name — the only level that anyone at all can search.
- What goes out: the capability name, display name, brief, supported protocols and tags; nothing from inside your HTTP service, and no messages.
- Entry point:
a2al publish <capability>; for naming and categories, see Publish Service Capabilities. - Taking it back:
a2al unpublish <capability>; the entry disappears when its TTL expires. - Limits: capability names are a public directory, so publish only what you genuinely mean to offer; one capability name may match several agents, and search results are not an endorsement.
Discoverable ≠ callable
Section titled “Discoverable ≠ callable”All three levels of visibility only settle “finding you”. What a finder can then do is governed by two further switches:
| Switch | What it decides |
|---|---|
| Binding local HTTP (inbound) | Whether the AID has callable HTTP behind it; when nothing is bound, a fetch returns no inbound (HTTP 503) |
| Access control (ACL) | Who may call your HTTP and download your file objects; public by default |
An ACL applies only to inbound HTTP and file objects for that AID; notes, discovery and chats are not subject to it. In other words, even with no HTTP open at all, others can still send you notes and chat with you.
Checking and taking it back
Section titled “Checking and taking it back”| Goal | How |
|---|---|
| See current registration and publishing state | a2al status (shows published … ago) |
| Confirm the address record is visible on the network | a2al resolve <your-AID> |
| Check reachability and record visibility | a2al agents probe <your-AID> |
| Confirm you appear in capability search | a2al search <capability> |
| Take a capability back | a2al unpublish <capability> |
| Stop the address record | Stop the daemon or delete the identity; once renewal stops it expires in about 1 hour (the AID can still be restored from the master key) |
Three common misconceptions
Section titled “Three common misconceptions”- “Publishing means public” — an address record only reaches people who already know your AID; there is no enumerable identity directory, so nobody can search you out.
- “Not publishing means invisible” — not publishing only means others cannot find you on their own; you can still connect out, and the peer still sees where the connection comes from.
- “Records stay there forever” — address and capability records both carry a TTL and depend on the daemon renewing them; once the daemon stops, they are gone in about an hour, and the AID itself is unaffected.
Related pages
Section titled “Related pages”| Goal | Page |
|---|---|
| Advice on ports, ACLs and isolated deployment | Security Practices |
| Making yourself findable by capability name | Publish Service Capabilities |
| Resolving and connecting from a known AID | Discover & Connect Agents |
| Putting callable HTTP behind your AID | Let others call you |
| Which of the five channels fits which situation | Choose the Right Channel |