Eve

Eve by Skytells is a master-class AI agent — orchestrator, creator, and operator. She coordinates multi-agent systems in real time, controls a live browser, executes account actions, and generates content on demand, all from a single conversation inside the Console.

Eve is Skytells' AI agent — and she is a powerhouse. Not just an agent: a full operational system capable of orchestrating multi-agent pipelines, controlling a real browser inside an isolated VM, generating content, and executing any Console operation with a single command.

Eve's multi-agent collaboration architecture was pioneered by Skytells long before multi-agent systems became an industry standard. What is now a headline feature across the AI ecosystem was already foundational to Eve's design from the beginning.

Eve's full product is available at skytells.ai/eve. Inside the Skytells Console, Eve is embedded as your operational AI companion — fluent in the platform, aware of your context, and capable of acting, not just advising.

What makes Eve different

Eve is not a chatbot. She is an agent with genuine operational capability:

Orchestration — Eve delegates subtasks to specialized agents in real time, synthesizing their outputs into a single coherent result. She manages task graphs, resolves dependencies, and handles failures without surfacing complexity to you.
Isolated VM execution — Eve runs human-level tasks inside an isolated real VM. Browser sessions, code execution, and system interactions happen in a sandboxed environment — contained, auditable, and safe.
Browser control — Eve operates a real browser inside that VM. She can navigate pages, make mouse movements, fill forms, click elements, and observe what appears on screen — executing workflows the way a human operator would.
Creative generation — Eve can generate images, compose and perform songs, produce audio, and create rich media assets — by invoking the appropriate Skytells models on your behalf.
Single-command Console operations — Everything that can be done in the Console can be done by Eve with one command. She is the operational layer for the entire platform.
Live app building — Given a goal, Eve can scaffold, configure, and deploy an entire application, coordinating the steps across the Console in real time.

Eve in the Console

Within the Console, Eve handles the operational layer — keeping the experience smooth and removing friction from complex workflows. She is always aware of the page you are on, the project you are working in, and your recent activity.

Access Eve via Eve in the left sidebar, or by pressing the keyboard shortcut shown in your account settings.

What Eve can do

Orchestrate agents

Eve coordinates multiple specialized AI agents in real time — assigning subtasks, managing dependencies, and synthesizing results. Ask her to accomplish complex, multi-step goals and she handles the full execution graph.

Control a real browser

Eve operates a live browser: navigating URLs, moving the mouse, clicking elements, scrolling, filling forms, and observing what renders on screen. She executes workflows with real browser sessions, not simulations.

Generate images

Ask Eve to create an image — she selects the right model, constructs the prompt, runs the prediction, and returns the result, all in the same conversation.

Compose and perform music

Ask Eve to write and generate a song. She will author lyrics, select a music model, and produce a full audio track — delivered directly in the conversation.

Navigate the Console

Eve knows the full structure of the Console. Ask her to go anywhere — by name, description, or intent — and she navigates directly.

Execute account actions

Create API keys, configure projects, set environment variables, manage deployments — Eve can take direct actions on your account.

Build apps live

Describe what you want to build and Eve can scaffold, configure, and deploy an entire application — working through the Console in real time on your behalf.

Surface predictions and history

Ask Eve to find a specific prediction, summarize your recent inference activity, or locate a model by capability.

Example prompts

Creative generation

Sing me a song about deploying to production at 3am.

Generate a cinematic image of a futuristic data center.

Compose a lo-fi hip-hop track, 30 seconds.

Create a product demo image for TrueFusion.

Browser & app building

Open my dashboard and summarize what you see.

Build me a simple Next.js app and deploy it to my project.

Go to my apps list and show me which ones are failing.

Take me to my predictions.

Open the Model Catalog.

Go to my API keys settings.

Navigate to the LLM Playground.

Explanation

What is the difference between an App and a Deployment?

How does zero data retention work on Skytells?

Explain the difference between Tokens and API keys.

What does the "Source" column mean in Predictions?

What is the Predictions API?

Workflows

Help me create an API key.

Guide me from Projects to deploying a Docker app.

Walk me from Billing to upgrading my plan.

How do I add a custom domain to my app?

I want to generate an image with TrueFusion. Where do I start?

Predictions

Show me my latest prediction.

View prediction 05bd8fe1-7646-4b3d-89d1-a0e2f2a3b4c5.

What did my most recent Flux 2 Flex prediction return?

How much did I spend on predictions today?

Models and Playgrounds

Find me a model that does text-to-video.

What image models are available from OpenAI?

Open the Inference Playground for TrueFusion Edge.

What is the price per image for TrueFusion Max?

Context awareness

Eve is always aware of the page you are currently viewing. This means your questions can be implicit — you do not always need to specify what you are looking at:

If you are on the Predictions list page and ask "What does the Source column mean?", Eve knows you are referring to the Predictions table.
If you are viewing a model card and ask "How much does this model cost per image?", Eve reads the displayed pricing.
If you are on the LLM Playground and ask "What should I set temperature to for creative writing?", Eve knows the context and gives model-appropriate advice.

This context awareness makes Eve significantly more useful than a generic chatbot or documentation search — it understands your position in the Console and tailors its response accordingly.

Conversation persistence

Every Eve session is a named conversation associated with your account. Conversations are preserved across browser sessions and Console logins.

You can return to a past conversation at:

console.skytells.ai/eve/{conversationId}

The conversation list is shown in the Eve sidebar on the left when you open the Eve interface. Each conversation is timestamped and labeled by its first message or topic.

To start a new conversation, click New conversation at the top of the Eve sidebar.

Conversation history is stored server-side and tied to your Skytells account. It persists across devices — start a conversation on your laptop and continue it on your phone.

Eve's capabilities in depth

Orchestration and multi-agent collaboration

Eve's defining capability is orchestration. When you give her a complex goal, she decomposes it into a directed task graph and assigns each node to the most suitable specialized agent — image generation, code execution, browser automation, API operations, or knowledge retrieval. Agents execute in parallel where dependencies allow. Eve monitors their progress, handles failures by rerouting or retrying, and assembles the final output.

This architecture was designed and deployed by Skytells before multi-agent collaboration became an industry standard. It is not a feature layered on top of a chat model — it is the foundation of how Eve operates.

Browser control

Eve operates a real browser with genuine interaction fidelity:

Navigates to any URL
Performs mouse movements, clicks, and scrolling
Fills forms and submits data
Reads and interprets what is rendered on screen
Takes screenshots and uses visual context to make decisions

This enables Eve to execute browser-based workflows — including third-party sites — not just actions within the Skytells Console.

Isolated VM execution

Every human-level task Eve performs runs inside an isolated real virtual machine. The VM is provisioned per session, sandboxed from other accounts and system resources, and terminated when the task completes. This means:

Browser sessions cannot persist or leak state across users
Code execution is fully contained
Actions are auditable and reproducible
The host system and your account data are never exposed to the task environment

The VM model is what allows Eve to safely perform real browser interactions, run scripts, and operate like a human operator without any of the security risks of doing so in a shared environment.

Tips for effective use

Be specific about what you want, not just what you want to know. "Help me deploy a Docker app" is more actionable than "What is a deployment?"
Provide context when you are troubleshooting. Include error messages, model names, or prediction IDs so Eve can give targeted help.
Ask follow-up questions. Eve retains context within a conversation, so you can say "Now do the same for the other environment" and it will understand.
Use Eve for discovery. If you are not sure where something is in the Console, ask Eve rather than clicking through menus.

Privacy and security

Eve operates with full access to your account context — the page you are viewing, your project context, prediction output data (images, video, audio files), and all resources your account has access to. Eve can read, display, and act on your prediction outputs as part of delivering results.

Eve does not have access to secret values (API key secrets, environment variable values), which are write-only and never surfaced — even to Eve.

All Eve conversations are stored encrypted, tied to your Skytells account, and not used for model training or shared with third parties.

Eve was designed and operates in full conformance with Skytells Responsible AI — covering safety, ethics, and governance frameworks that govern how Eve reasons, acts, and declines requests.

Eve product page — the full Eve product outside the Console.
Responsible AI — the safety, ethics, and governance frameworks Eve operates under.
Predictions — the predictions list Eve can navigate to and surface records from.
Model Catalog — Eve can find and navigate to specific models.
Inference Playground — Eve can open the Playground for any model you describe.

How is this guide?