Known Issues
As of v3.5.5 (Aug 17, 2026), Memo's internal bug tracker shows zero open bugs at every severity — every previously-tracked issue has been fixed and verified with a regression test. What's below are genuine, currently-open limitations of features still in beta (including self-hosting), not bugs.
Self-hosting limitations
No built-in TLS yet
The backend serves plain HTTP. For anything reachable outside your own LAN, put it behind Tailscale, ngrok, or your own reverse proxy (nginx/Caddy) with TLS termination — don't expose port 8090 directly to the internet.
Tailscale/ngrok tunnels are Settings-only from the CLI's perspective
memo remote/memo service manage auth and the systemd unit, but starting or stopping a Tailscale/ngrok tunnel itself is still a desktop-Settings action, not yet a CLI subcommand.
No adversarial security test against a real deployment yet
The 4-mode auth system, per-device tokens, and CORS fix have all been functionally tested (including live, against a real non-loopback client on a Raspberry Pi), but a genuine adversarial penetration test of a real self-hosted deployment hasn't been done yet — treat it as beta-hardened, not audited.
Orchestra + Agent Mode: permission dialog can lag a finished-looking chat turn
With Orchestra and Agent Mode both on, the live chat view can occasionally show a turn as already finished while an Orchestra task's tool-permission request is still resolving in the background. The backend itself handles this correctly (the real 60-second timeout still fires and the result is saved) — it's a display-timing gap, not a lost request.
Beta feature limitations
Live Mode — no echo cancellation yet
Voice chat has no acoustic echo cancellation, so using speakers instead of headphones can occasionally make Memo mistake its own spoken reply for you interrupting it. Full duplex audio is planned for a later release.
Memo Swarm — not available on macOS
The multi-PC compute-pooling helper binary isn't packaged for macOS yet; Swarm currently requires Linux or Windows machines.
Claude Code / Codex CLI providers — no review UI
There's no interface yet for reviewing what a CLI-backed chat agent actually did (file edits, commands run) beyond its final text reply.
Developer API Gateway — partial tool-calling support
Full agentic tool calling works for the openai/custom/local/groq/openrouter/grok/opencode-zen/opencode-go providers. It's not yet available for gemini/claude/ollama, whose own provider implementations don't support tools — a tools-bearing request to one of those returns a clear error instead of silently dropping the tools.
Planned work
- Test coverage for previously untested areas:
handlers_oauth.go,handlers_proactive.go,internal/cloudsync/drive.go,hardwareID(). - Deep bug-scan pass over modules not yet covered by the multi-agent review done for v3.3.3:
internal/cloudsync,internal/skill,internal/proactive,internal/observer. - Adversarial security test of a real self-hosted deployment (see above).
For everything previously listed here and since fixed, see Resolved Issues.