Guide

Getting started

From a fresh install to your AI co-host speaking in a live room — in about five minutes.

What you'll need: the Maneku desktop app, a Maneku account, and a recording room link. The rest depends on your plan — see below.

Two plans — pick yours: On a credit plan, you buy credits and we run the AI for you — no API keys to manage. Skip to step 3. On a bring-your-own-key (BYOK) plan, you connect your own AI accounts and pay those providers directly — do step 2 first.

1. Install, sign in & first run

Download Maneku from maneku.ai/beta (during the beta) or maneku.ai/download, open the file, and drag Maneku into Applications. On first launch, create your account:

On a credit plan you're ready to go to step 3. On a BYOK plan, the app asks for your API keys next.

Maneku's first-launch key screen
The first-launch key screen (BYOK plans). Anthropic and Groq unlock the app; ElevenLabs is optional.

2. Add your API keys BYOK plan only

On BYOK you connect your own AI accounts, and usage is billed by those providers directly. Keys are stored encrypted on your device, sealed with a key held in your macOS Keychain; they're never sent to Maneku. On a credit plan there are no keys to add — skip to step 3.

ProviderPowersKey looks like
AnthropicThe co-host's responsessk-ant-…Required
GroqLive transcript the co-host responds togsk_…Required
ElevenLabsHigher-quality voice (vs. system voice)sk_…Optional

Paste each key into its field and hit the save arrow; Maneku stores it and checks it against the provider. You can add or change keys anytime from Settings (the gear icon). Anthropic and Groq are both required and the app won't open past the key screen without them — Anthropic generates responses, and Groq produces the live transcript the co-host listens to, so without it the co-host has nothing to respond to. ElevenLabs is optional; without it, Maneku uses your system voice.

Need the click-by-click? The key-setup walkthrough shows exactly how to get each key from Anthropic, Groq, and ElevenLabs (including the gotchas). If a key shows an error, re-check it in Settings — most problems are a copy/paste slip or an expired key; see Troubleshooting.

3. Create a show & pick a persona

Hit Start a show on the home screen. A show is a preset for a podcast or series — its name, default persona, and any standing context. Create one from the Show dropdown:

The New show editor with name, persona, theme, and pre-show brief fields
The show editor. The pre-show brief is standing context the co-host gets for every episode of this show.

A persona is the AI character that joins: its name, voice, and personality (you can tune humor, formality, and how chatty it is). Pick one from the Joining As dropdown — Co-Host, Skeptic, and Researcher come built in — or create your own:

The persona editor with personality text, sliders for humor, formality and verbosity, response length, model and voice choices
The persona editor. Personality text plus sliders, response length, model, and voice.

4. Point it at your room & join

On the pre-connect screen, paste your room link into the URL field. Maneku detects the platform automatically and runs the right join flow.

The pre-connect screen with a show and persona selected and the room URL field
Ready to join: show picked, persona picked, room link goes in the URL field.
PlatformStatus
DescriptSupported
RiversideSupported
RestreamSupported
DiscordExperimental

Optionally add episode notes (free-form context sent with every summon) and reference materials — paste an article or video link and Maneku pulls in the text so the co-host can draw on it. Then click Join Room. You'll see the status move from loading to filling in the name to joined; once it's in, the panel flips to the in-room view with your stage and transcript.

Episode notes filled in on the pre-connect screen
Episode notes: tonight's context, sent along with every summon.
The reference material modal with a pasted show-notes source and a fetched Wikipedia article
Reference material: paste text or a link and Maneku fetches it. Attribution controls how the co-host credits each source.

Before you record: let the room know an AI co-host is present and that you're recording. We've written ready-to-use scripts in the AI Disclosure & Consent Guide.

5. Summon the co-host

The co-host listens but stays quiet until you bring it in. When you want it to speak, summon it. There are three types:

TypeWhat it doesRelative cost
AskA quick, lightweight reply to what's being saidLowest
ResearchCan search the web to back up its answerMedium
DeepA longer, more considered responseHighest

Trigger a summon three ways:

The co-host thinks, then speaks into the room. To cut it off mid-sentence, press Esc or hit Stop. A live counter in the header tracks summons and estimated cost for the session, and you can set budget alerts in Settings.

The in-room view: live transcript on the left, the speaking co-host with summon buttons and stop control on the right
Mid-summon: live transcript on the left; the stage with Ask / Research / Deep, the response, and the Stop button on the right.

6. After the show

Your transcript and a session summary (duration, summon count, estimated cost) are saved on your device, and you can export the transcript from the app. Your personas, shows, and settings sync to your account so they follow you across devices. Transcripts do not sync — they stay on your machine. The full picture of what stays local and what doesn't is at Where your data goes.

The home screen showing a recent show card with its summary
Back home after the show: the session card with duration, summon count, and an AI recap of the episode.

Stuck on something? Head to Troubleshooting for fixes to the most common issues, or reach us at support@maneku.ai.