Less chasing.
More doing.

Mendy said no: voice calls use 0.2 message credits per minute as their charge, rather than another per-message fee on top. He distinguished that from meeting note-taking, which he described as not charged. These are session-time statements, not a current pricing guarantee.

Mendy described an upcoming owner-controlled agent or subagent flow with its own WhatsApp number. He mentioned Meta approval typically taking around 24 hours. He did not confirm arbitrary multi-number behaviour on an existing Superagent as a shipped feature.

It was rolling out. Mendy offered to enable access for attendees who supplied their account email privately through Sam; this recap does not expose those emails or imply every account was enabled.

Mendy thought builder support would come in the future. No date or completed builder implementation was announced.

No, not in the feature shown. Mendy said it was purely a note-taker for now.

Mendy previewed a temporary, one-use card issued through a Link-based flow after the user reviewed merchant information and approved the transaction. No live payment was made, and broad autonomous spending was not presented as the permission model.

It belongs to the workspace and can be assigned to an app, but it does not need to work on an app at all. A personal assistant is also a valid use case.

Sam pointed to earlier end-to-end Office Hours, community voice or chat help, and paid partner help for a more involved build. The reservation flow itself was not implemented during this session.

Sam said onboarding was being developed to clarify the assistant's intended role and app connection before entering the conversation. He showed a newer onboarding flow but noted viewers might see a simpler one.

Sam described a goal as an objective or end state pursued over multiple steps and time, rather than a single immediate request. Mendy's slide-production example included gathering information, delegating, waiting, and handing work to people.

Sam said this could be built through a phone app connected to the agent. He did not demonstrate a native face-to-face recording feature; capture, consent, and integration would need their own setup.

Yes. Sam pointed to the top-left dropdown or the Agent tab's create flow and later said there was no count limit in his explanation. The session did not audit plan-specific current quotas.

The live YouTube link persists as a VOD and can be bookmarked or rewound. Sam offered specific links while acknowledging that the unofficial recap archive was behind.

Sam suggested providing the other agent's ID or using a purpose-built app as an intermediary. No secure communication implementation or loop-control test was completed here, so access, credentials, and budgets still need explicit design.

Sam said a natural-language request such as size, price ceiling, and comfort could guide the research. He did not demonstrate a dedicated lower-cost semantic-search backend or validate product results in the session.

Sam gave examples of small-business phone operations feeding a CRM, a family meal-planning experiment, and multi-agent business intelligence across marketing, recruitment, competitors, and market gaps. These were anecdotes rather than audited case studies.

Sam said the agent's phone number could be given directly or used with a provider's forwarding feature. The exact trusted-caller policy was later left for Mendy to clarify.

Sam said yes and cited the community Hivemind project as an example. The live research demo also showed one assistant delegating to three subagents; these are not necessarily the same as independently configured workspace Superagents.

Sam emphasised the tradeoff between capability and cost, with personal preferences for expensive models on demanding work. He recommended choosing what works for the user's budget rather than providing a measured universal model ranking.

Sam speculated that browser control might allow it, but recommended waiting for the official approved-payment method Mendy had described. No Google Pay capability was tested.

Send evidence to support. Sam said a confirmed bug could justify a credit correction, but no account investigation or refund occurred live and eligibility is not guaranteed by this recap.

Create a Superagent, confirm the target app, give the market and business objective, ask for sourced competitor and pricing research, then inspect the results. Sam demonstrated this with a fictional Lake District dog-walking business.

They are workers the Superagent spins up to handle parts of its task. The demo used three parallel research assignments before producing a consolidated report.

No. Sam repeatedly distinguished Dashboard → Agents inside an app from the workspace Superagent created through the home-page Agent tab. Do not assume the former has all the calling, research, or orchestration features discussed here.

No supported pathway was confirmed. Sam suggested experimenting with an exported configuration or asking a Superagent to inspect the app, and explicitly said he had not tested it. Entity-stored information may be accessible with permission; chat memory was not shown to migrate.

Sam said no separate update was needed: use the Agent tab on the Base44 home page. Individual new features could still be rolling out.

Sam and the audience discussed an experimental Codex/API setup, but no procedure or working connection was demonstrated. Such access would be powerful and potentially costly, so the recap does not treat a pasted API key as a verified or safe one-step setup.

Sam did not know. He guessed around six credits per day but explicitly said the first run would need to be measured. That guess is not a usage estimate to rely on; model choice and the work performed affect cost.

Sam said that experiment could consume both Base44 credits and OpenAI usage. No rates or total were measured in the session.

The research report appeared and Sam requested daily delivery, then said the automation was created. He did not connect WhatsApp on stream or show an actual scheduled message arriving, so end-to-end delivery remains unverified in the recording.

Open a support ticket at app.base44.com/support. Community get-help channels and Wednesday Office Hours can provide assistance, but Sam clarified that community management is not the dedicated support team.

Mendy's tour mentioned trusted callers, but Sam could not confirm the detailed policy when asked later and said he would check with Mendy. This remains an explicit follow-up question.

Sam asked whether the viewer was on mobile and suggested checking from a laptop or desktop. He suspected a rollout difference but did not confirm the cause or enable the feature live for that user.

In the handover to Kate, Sam explained that the screen-share dialog offers Applications for a window, Entire screen for a monitor, and Devices. This was setup help for the next session, not part of its LMS build.