
Meet the AIGE Agent: Your In-App AI Teammate
You open a generation tool, fill in a form, wait, tweak, repeat. That's fine for one image. It gets tedious fast when what you actually want is "make me ten ad variations for this product" or "turn this into a week of posts." That's a whole small project, not a single click.
That's what the AIGE Agent is for. It's a chat-first AI teammate built into the studio at aige.ws/agent. You describe what you want in plain language, and it plans the steps, calls the right generation tools, and hands back results in the same conversation. No separate app, no external client to configure. It's already part of AIGE.
What it actually is (and isn't)
Worth being precise here, because AIGE has two different ways an AI can drive generation, and they solve different problems. This article is about the in-app agent at /agent, a page inside AIGE itself, with its own chat thread, projects, and panels. If you instead want to drive AIGE's image and video generation from an external AI client (Claude Desktop, Claude Code, Cursor), that's a different feature, AIGE's MCP server, covered in connecting Claude to AIGE via MCP. Same underlying generation power, two different front doors. This one lives where you already work.
Under the hood, the agent runs on gemini-3.5-flash with a thinking-level toggle: a fast "minimal thinking" mode for quick turns and a "high thinking" mode for anything that benefits from more deliberation before it acts. You don't manage this directly beyond picking a mode; it's there so simple requests stay snappy and complex multi-step ones get more reasoning behind them.
Projects and sessions, not one long scrollback
The agent organizes work into projects, each holding its own sessions, so a product-launch project and a personal-character project don't get tangled in the same thread. Every session remembers its own conversation, and you can jump between them from the sidebar. Press ⌘K (or Ctrl+K) anywhere on the page to pull up search across your sessions instead of scrolling to find one.
Four dedicated views: Files, Memory, Connectors, Skills
Alongside the chat, the agent page has four other views you can switch to from the sidebar. Each one replaces the chat rather than sitting beside it, so you're either in conversation or in a panel, not both at once:
Files lets you upload assets to a project (images, references, documents); they stay attached to it, ready to reference in any session without re-uploading.
Memory holds durable facts and preferences the agent has saved about you across sessions (via a dedicated
remembertool), plus pinned assets you've told it to keep on hand for future turns.Connectors lists external services you can link, including Google Drive, each with its own connect/disconnect toggle so the agent can pull from services you already use.
Skills is a library of instruction sets the agent can apply to a request. AIGE ships dozens of built-in skills covering platform formats (Instagram carousels, TikTok trends, YouTube thumbnails, LinkedIn carousels), visual styles (film noir, golden hour, watercolor, anime cel), and model-specific techniques. The agent calls one when it clearly matches what you asked for, and can compose more than one per turn.
Quick-start recipes if you don't know where to begin
New session, blank composer, not sure what to type? The empty state offers a small set of recipe chips to start from instead of a blank page:
Social Pack: a multi-format set of ad images across platforms like Instagram, TikTok, and Pinterest, varying hook angle and framing per platform.
Cinematic: a short, story-driven video shot with mood and camera direction built into the prompt.
Product: studio-grade product photography from a URL or an attached photo, in multiple styles (lifestyle, studio, detail, flat-lay).
Custom: a blank canvas, for anything that doesn't fit a template.
Tap one and it pre-fills the composer with an editable template. Swap the bracketed placeholders for your specifics and send. First time on the page, a short onboarding tour walks through the layout before you start; you can skip it, and it won't reappear once you have.
What it can actually do when you send a message
This isn't a chatbot that just talks about generating. It calls real tools mid-conversation. Confirmed capabilities include generating single images or a whole batch of prompts in parallel, generating video from a starting image, checking on a running generation by task or batch id, and expanding a short prompt into a richer one before it generates. It can create and manage persistent characters, including creating a brand-new one on the spot when you say "name her X" from an attached photo, so a face it generates today can be reused in tomorrow's session. See how persistent characters hold a face across generations for how that identity-anchoring actually works.
Beyond generation, it can check your credit balance and plan before it spends anything, pull your recent generation history, publish a result straight to the AIGORA feed, read a web page you paste in (pulling headings, text, and image URLs, handy for that Product recipe), and run live web searches, including up to 8 in parallel for comparison research. It also has utility tools most chat assistants skip entirely: precise arithmetic, date math, JSON/CSV/Markdown table conversion, and entity extraction (pulling emails, URLs, or hashtags out of pasted text).
For multi-step requests, it drafts an explicit ordered plan before executing rather than improvising silently. You see the steps it committed to, not just the outcome. And it can schedule itself: give it an explicit future or recurring time ("every Monday at 9am, draft three post ideas") and it creates a scheduled task that fires a fresh agent turn later, which you can list or cancel afterward.
Honest note: the agent only schedules when you give it explicit timing. Say "generate an image" with no time attached and it generates now. It won't quietly queue a normal request for later, and it won't invent a schedule you didn't ask for.
Quick answers
Question | Answer |
|---|---|
Is this the same as the Claude MCP integration? | No. The agent at |
What model powers the agent? |
|
Can it remember things between sessions? | Yes. The Memory panel holds durable facts and pinned files the agent has saved, and each project keeps its own Files. |
Does it just chat, or does it actually generate? | It calls real generation tools (single images, batches, and video) plus utilities like web search, page reading, and scheduling. |
Where do I start if I don't know what to ask? | Use one of the recipe chips (Social Pack, Cinematic, Product) on the empty state; they pre-fill an editable prompt template. |
Can I connect outside services? | Yes, from the Connectors panel. Google Drive is one of the available connectors, each with its own connect/disconnect toggle. |
Do this next
Open aige.ws/agent, tap the Social Pack or Product recipe chip, swap in your own details, and send it. You'll see the plan it commits to and the first results in the same thread. No separate tool, no new tab.



Comments
Be the first to comment.