What happened
Krea presents Bots as AI teammates with a face, a job and a memory. The page lists roles including brand designer, character designer, image maker, retoucher, social manager, illustrator, trend researcher, colorist, copywriter, motion designer, set designer, packaging designer, 3D artist, creative director, stylist, video editor, storyboard artist, sound designer and layout designer. It says each bot performs one craft; users write its instructions once, and it follows them each time.
The collaboration model centers on chats. A person can mention a bot by name with @ and hand it a brief. The bot asks questions when something is unclear, completes its part, and can pass the next step to another bot. Krea describes a first bot, the director, that plans work, makes a first pass and brings in the bots it needs. The page illustrates a launch workflow involving image creation, storyboarding, 3D, motion, sound, video editing, social posts and brand design, but these are presented as examples rather than independently measured results.
Krea describes each bot as having a model, its own computer, connected apps, a working brief and tools. The example integrations and capabilities include Slack, Meta Ads, Krea’s editing, motion, sound and 3D tools, Figma and a browser. The page says a bot can build campaigns and launch them when approved. It also says the bot can inspect the images and cuts it makes and the references it receives.
The memory model has three parts: lessons about a bot’s craft learned from feedback; notes pinned to a project chat and available to every bot in that chat; and workspace-wide house style followed across chats. Krea says these are read before work begins. Lessons learned by a teammate wait for the user’s approval, with options to keep, drop or share them. The page invites visitors to join the beta waitlist and advertises $1,000 in credits; it does not state the credit terms or a general availability date.
What the archive adds
New Runtime’s earlier coverage has described both cross-device agent identity and team bot runtimes, but Krea’s announcement is a new event rather than a restatement: its focus is a set of named creative crafts and reusable creative memory. Relevant earlier notes include nanoMuse’s shared identity across phone, desktop and web (October 11, 2026) and Kwant’s self-hosted team bots with handoffs, routines and approvals (October 24, 2026).
Trend signal
Krea’s craft-specific roles, project and workspace memory, and chat handoffs support the idea that agent products are becoming composable capability surfaces rather than single general assistants. This fits agent-ready software exposes capabilities as a product-design signal, though Krea’s page does not establish how reliably those capabilities work in production. The specific memory design also raises a governance question: who can teach a bot, and who approves what it retains?
Why now
Krea is recruiting beta users and attaches a $1,000 credit offer to the waitlist invitation. That makes this a product-access announcement, but the page gives no beta start date, qualification criteria or credit redemption details.
Connections to other events
Krea’s shared-chat handoffs meet a broader product pattern also visible in New Runtime’s earlier coverage of Kwant’s team-bot runtime, which described bot handoffs, routines and approvals. Krea applies a related coordination idea to creative production, with a director assigning work among specialist roles rather than presenting the bots as general-purpose coworkers.
Why it matters
For a creative team, the interesting unit is not a single image generated from a prompt but a sequence of bounded jobs: plan, create, revise, package and distribute. Krea’s design makes those jobs explicit as roles, so a team could give each one a stable brief and connect handoffs inside a shared conversation. Project notes and workspace style could reduce repeated briefing, while craft-specific feedback could make a role more consistent over time.
The same structure creates operational questions teams should resolve before relying on it. A persistent bot needs a clear owner, a review point before consequential actions, and a way to inspect what it has learned. Krea says teammate lessons await approval and Meta Ads campaigns launch only when approved, but it does not describe a full permission model or audit trail. Teams should assess output quality and human review effort per workflow, rather than treating the page’s illustrative schedule as demonstrated productivity.
Limits and uncertainty
Krea’s page describes the product and its intended workflow, but does not provide independent evaluations of output quality, reliability, handoff success or time saved. The hours of bot work and minutes of human attention shown in the example schedule are illustrative claims, not a measured customer study. The page does not specify beta eligibility, the $1,000 credit’s terms, ongoing pricing, supported limits, or when the product will become generally available. It also does not detail permissions, logging, data retention or how bot memory can be exported or deleted beyond the stated keep, drop and share controls for lessons.
Skill to build now
Map one repeatable creative workflow into distinct jobs before trying a multi-agent setup: write each role’s input, output, acceptance check and handoff condition. Separate project-specific notes from reusable house style, and decide who approves changes to the shared instructions. This gives a team a concrete test plan for Krea Bots or any similar system, without assuming that role labels alone make the work dependable.
Reader takeaway
Krea Bots package creative agents as persistent specialists that can coordinate in chats and draw on layered memory. The promise is less repeated briefing and more structured handoffs; the practical test is whether a team can verify each role’s output, control what it remembers and get reliable results at a known price.
Open question
Can Krea show that these bot handoffs reduce total review and coordination time on real creative projects, not just the time humans spend messaging agents?