Getting help

Mel is in active development, which means two things: things occasionally break, and your reports genuinely shape what gets fixed next. Here's how to reach us and how to make a report we can act on fast.

Contact

Use the contact form — messages go straight to our support inbox, and replying reaches you directly. Prefer email? info@helloagentic.ai is the same inbox the form delivers to. Either channel is right for everything: bug reports, questions, feedback, and account issues.

Accounts

The Mel agent needs a free Mel account, and sign-up is self-serve — there's no waitlist. Create one from the sign-in panel in the app (it's instant), and you're ready to use the agent. If a sign-in or account issue is blocking you, email us with the address you signed up with.

The terminal itself — blocks, panes, tabs, the editor, code review — works without an account. Only the agent needs sign-in.

Reporting a bug

A report we can reproduce is a report we can fix. Please include:

• Your OS and version — e.g. macOS 15.2, Windows 11, Ubuntu 24.04.
• Your Mel version — from the About settings, or match it against the changelog.
• What you did — the exact command, prompt, or click sequence. "The agent broke" is hard to chase; "I ran /orchestrate with this prompt and the second worker pane never started" is gold.
• What happened vs. what you expected — including the exact error text, copied verbatim. Relay errors have stable machine codes (unauthorized, forbidden, payload_too_large, …) that tell us precisely which path failed.
• Agent context, if it's an agent bug — which model the chip showed (e.g. "auto → gemini-2.5-flash"), whether the footer showed a relay offline or connecting… pill, and roughly when it happened (helps us find it in server logs).

Before reporting a connectivity issue, run through the Agent connectivity checklist — a cold-start pause or a proxy blocking streamed responses accounts for a lot of "the agent is down" reports.

Before you write in

A quick self-serve pass often resolves things faster:

1. First-launch or install trouble → Install issues.
2. An error message → Common errors — most of the frequent ones are decoded there.
3. Behavior that seems wrong → check the changelog; it may have changed on purpose in a recent release.
4. "How do I…" questions → the feature docs, starting from the Introduction.

Feedback that isn't a bug

Feature requests, rough edges, "this felt confusing" — all welcome at the same address. We read everything, and small specific observations ("the filter bar hides the match count when the pane is narrow") are exactly the kind of thing that ships in the next release.