feat: Tailscale serve integration for secure remote access

[teknium1]

Feature Request

Source: Community request from @PaulTisl

Problem

Users want secure remote access to the Hermes API server / Open WebUI without exposing ports publicly or setting up reverse proxies. Tailscale's tailscale serve provides zero-config HTTPS tunneling over a private mesh network — Claw3D already uses this pattern.

Proposal

Add optional Tailscale serve integration so users can expose their local Hermes API server over their Tailnet with a single config toggle or CLI flag.

Possible approaches (lightest to heaviest):

1. Docs only — Document the one-liner: tailscale serve https://localhost:8642 alongside gateway setup. Zero code changes.

2. Config flag + auto-launch — Add a gateway.tailscale_serve: true config option. When enabled, hermes gateway start spawns tailscale serve pointing at the API server port as a subprocess alongside the gateway. Teardown on hermes gateway stop.

3. CLI flag — hermes gateway start --tailscale as a one-shot option without persisting to config.

Considerations

• Tailscale must already be installed and authenticated (tailscale status should succeed)
• The API server key (HERMES_API_KEY) should be enforced when exposing remotely — warn if unset
• tailscale serve handles HTTPS termination automatically
• launchd/systemd service definition may need updating if we want Tailscale serve to persist across reboots alongside the gateway

Prior Art

• Claw3D uses tailscale serve for remote access (referenced by the requester)
• Only one Tailscale reference currently exists in the codebase (URL safety test)

[GottZ]

This was generated by AI during triage.

Summary

One PR addresses #9269. #9295 implements the requested Tailscale Serve path through persistent configuration and a one-shot CLI flag, including readiness checks, API-server lifecycle integration, teardown, tests, and documentation, but it does not cover the broader Funnel mode requested in duplicate issue #8951.

Related pull requests

• feat: Tailscale serve integration for secure remote access #9295 [closed] best fix — (+441/-12) — partial fix, closed but still the best existing fix and reference implementation: the diff directly adds gateway.tailscale_serve, hermes gateway run --tailscale, Tailscale status validation, Serve startup and teardown, API-key warnings, tests, and documentation; it also contains unrelated quick_test.log and test_failures.txt artifacts that should be removed from any successor.

Suggested consolidation

#9295 is the best existing fix, but it is closed and should not be advanced as-is. Author action: rebase onto main, or split out the part that can merge—preserve the tested Serve lifecycle, configuration/CLI wiring, and documentation while dropping the two test artifacts and explicitly scoping the result to Serve rather than the broader Serve/Funnel request in #8951.

Complex graph

flowchart LR
    classDef open fill:#dbeafe,stroke:#1d4ed8,color:#1e3a8a
    classDef merged fill:#dcfce7,stroke:#15803d,color:#14532d
    classDef closed fill:#e5e7eb,stroke:#6b7280,color:#1f2937
    classDef unverified fill:#f3f4f6,stroke:#9ca3af,color:#374151
    classDef best stroke-width:3px,stroke:#b45309
    classDef target stroke-width:3px,stroke:#4338ca
    I8951(["issue #8951 (open)"])
    I9269(["issue #9269 (open)"])
    P9295["PR #9295 (closed)"]
    P9295 -->|best fix| I9269
    P9295 -->|best fix| I8951
    class I8951 open
    class I9269 open
    class I9269 target
    class P9295 closed
    class P9295 best
    class P9295 best
    click I8951 "https://github.com/NousResearch/hermes-agent/issues/8951"
    click I9269 "https://github.com/NousResearch/hermes-agent/issues/9269"
    click P9295 "https://github.com/NousResearch/hermes-agent/pull/9295"

Loading

Graph: solid arrow = fixes / best fix, dashed arrow = partial or unverified (see edge label); boxed group = PRs duplicating each other; amber border = best fix; indigo border = target; gray node = closed (state tag in the node label).

Cross-PR triage: Reviewed 1 pull request and 2 issues in this complex. Each diff was read against this issue; Assessment working set: 30 kB of PR diffs, 6 kB of issue/PR text, 4 kB of discussion (6 comments), 4 verify verdicts. verdicts reflect diff content, not PR titles. Part of an automated triage batch.