NetShow
Back to Help
Guide

Create an agent

4 min read

The creation flow is at Dashboard → Create your agent (/dashboard/agents/create-stitch). It is five steps in one page. Each step opens the next when it has what it needs, and the panel on the left previews the agent as you go, so you are never guessing what you are making.

You need to be signed in. Nothing here is billed, and nothing is written until the last button.

1 · Identity

Three fields.

  • Name — what the agent is called. This is also what its web address is built from later.
  • Tagline (optional) — one line on what it is here to do.
  • Description — what it does, and for whom.

Fill in the name and the description and the step's button turns on. It is labelled Give it a face.

2 · Its face

Two lanes, shown as tabs: Pick a face and Generate a face.

Picking is the short road. The gallery holds the estate's ready-made face bases — Sol, Dean, Volt, Marshall, Quinn — each with its role under the name ("solar tech · SunRay", "home inspector · West Coast"). Click one and it is selected; the preview panel starts wearing it immediately.

Generating builds a face from the agent's name instead. It is deterministic — the same name gives the same face — so it is a real identity rather than a random draw.

Either way the face previews live and free, and the step's button reads Bring it alive.

3 · Presence and voice

Presence tier is how the agent is rendered wherever it appears. Six cards, and they are not all open at create time — the cards say so themselves:

Tier What it is At create time
SVG A procedural vector face generated in your browser Open — previews live
Hybrid A real face base with a living overlay on top; eyes, brows and mouth move with the conversation Open — previews live
Live VRM A rigged 3D model with a live face plate — the deepest tier Open — you supply the model link, or use the starter model
Kinetic Typography The transcript itself is the face — the agent's words are its presence Open — previews live, with a starter face offered
ASCII Hand-authored character art Studio tier — the card is disabled here
Sketch Line A living line drawing Studio tier — the card is disabled here

The two studio tiers are not missing. Their own wording is "craft it in the studios after your agent exists" — they are authored in the studios rather than picked from a menu, so they are not part of a first-run.

Picking Live VRM reveals a model URL field and a Use the starter model button, which fills the field with the estate's own starter model. Kinetic Typography offers a starter the same way — Ticker, the estate's own kinetic starter face — because its face needs styling rather than art.

Voice is a short list of named voices. Choosing one binds it to the agent now; it is part of the presence identity that gets written in step 5.

The step's button reads Give it capabilities.

4 · Instructions and capabilities

Instruction set — three sources, one of which is marked Recommended:

  • Persona blueprint (recommended) — compiled from the name, description and personality you already typed. The platform writes the tool and memory sections itself.
  • Start from a template — a proven role body you can edit after creation.
  • Write your own — paste or type the instruction body.

Whichever you pick, it is saved as version 1, and the page is explicit that versions only grow — nothing is overwritten later.

Capabilities are grouped: Knowledge & memory, Communication, Creation, and Commerce & wallet (marked Advanced, and off by default — the group's own note says commerce and wallet capabilities must be enabled per agent deliberately). Three are pre-selected for you: Web search, Memory recall, Memory store.

The step's button reads Review & create.

5 · Review, then create

The last step is a checklist of what you just built — identity, face, presence tier, voice, instruction set, capabilities — each with the value you chose beside it.

Under it, the page tells you exactly what pressing the button writes:

the agent record · its settings (capabilities, persona provenance) · alive profile v1 (face, tier, voice — the presence identity) · instruction set "primary" v1. Nothing else.

There is also a See the avatar package draft toggle if you want to read the presence record before it exists.

Tick the confirmation line, press Create agent, and the page hands you straight to your new agent's test drive screen (/dashboard/agents/{id}/test-drive) so the first thing you do is talk to it.

What happens next

Three roads open from there, and each has its own guide:

Still stuck?

The longer documentation covers models, voice, knowledge, and workflows in depth.

Open the documentation