Changelog

Skytells Cloud Storage

Cloud Storage is a cloud drive built for the Skytells ecosystem — secure, private, and integrated by design. Every workspace gets its own storage layer that lives on the same private network as your applications and services, with zero public exposure unless you explicitly choose it.

• Attach images, datasets, or any asset directly to a model inference request — no intermediate download or upload required
• Save model outputs, generated assets, and pipeline artefacts back to the drive without ever leaving the platform
• Accessible from every Skytells product — Playground, Deployments, Databases, Cloud Agents — because it shares the same internal fabric
• Optionally expose individual buckets or files to the internet with a single toggle; everything else stays air-gapped inside your project network

Cloud Storage is not a bolt-on. It is a first-class primitive that makes the rest of the ecosystem meaningfully more capable.

Project Tags

As projects grow, organisation becomes its own discipline. Tags give you a lightweight, flexible layer to bring structure to any number of projects without imposing a rigid hierarchy.

• Apply one or many tags to any project — production, staging, experimental, team names, verticals, whatever maps to your workflow
• Filter and group your project list instantly by tag
• Tags are shared across your organisation, so conventions stay consistent across teams

A small addition with compounding value as your footprint on the platform grows.

Write with AI

Crafting precise prompts takes real effort — especially when working across models with different capabilities, context windows, and optimal instruction styles. Write with AI is a built-in assistant that helps you go from intent to a well-formed prompt, right inside the Playground and anywhere else prompts are authored.

• Describe what you want in plain language; Write with AI structures and expands it into a prompt tuned for the selected model
• Adapts its suggestions to the active model's strengths and known behaviour
• Works across all supported providers — no context switching, no copy-pasting between tools

The goal is to reduce the overhead between having an idea and testing it, so iteration stays fast.

Usage Billing — Pay As You Go

Plan caps exist for predictability, not as hard ceilings that stop work mid-flight. If your usage exceeds your current plan's included quota, you can now enable Pay As You Go and continue without interruption.

• Activate from your billing settings in seconds — no plan change, no support ticket
• Usage beyond your cap is metered and billed at published rates, with a clear breakdown in your invoice
• Set a monthly spend cap so there are no surprises at the end of the billing period

For teams whose usage is inherently variable, Pay As You Go removes the friction of planning ahead and lets the platform scale to match actual demand.

At Skytells, every system we build is designed to work inside a single, coherent ecosystem — not as a standalone feature, but as a component that makes every other component sharper. Today's release reflects that directly.

Cognition is now integrated into Deployments

When a deployment runs, Cognition analyses the risk surface, parses deployment logs, and surfaces signals worth acting on — before they compound into incidents. Deployments have always given you visibility into what shipped. With Cognition, you also get a read on what that deployment means operationally:

• Anomalies in log patterns detected automatically
• Configuration drift flagged inline within the deployment view
• Risk indicators evaluated as part of the pipeline — no extra setup, no separate dashboard

New database engine: LibSQL

Skytells is the first infrastructure platform to offer instant LibSQL deployments alongside the full suite of supported engines: MariaDB, Redis, PostgreSQL, MySQL, and MongoDB. Every database engine can now be provisioned entirely off the public internet — living inside the project network and communicating directly with applications over the internal fabric.

• No public endpoint exposed
• No egress to manage
• Full isolation by default across all supported engines

Cloud Agents now support MCP via Trusted Gateway

Cloud Agents can now call MCP-compatible servers — internal or third-party — through the Skytells Trusted MCP Gateway, inside the same secure execution context they already run in. The gateway handles authentication, auditing, and scope enforcement, so agent capabilities expand without expanding your attack surface.

These three changes are not independent. An agent can trigger a deployment, Cognition evaluates what that deployment surfaces, and the database it reads from never touches public infrastructure. That is the ecosystem model we are building toward.

Today we are opening the doors to the Skytells Console. What was once reserved for enterprise organizations is now in the hands of every team building with intelligence at the core. The Console brings together application deployment, cloud infrastructure management, and access to over 50 foundation models from providers like OpenAI, Anthropic, Google, Mistral, and Meta into a single, thoughtfully crafted workspace. Every interaction has been designed around speed and clarity, because the tools teams rely on daily should never get in the way. We built Skytells with a clear conviction: the standard for how teams ship, observe, and scale intelligent software needs to be raised. The Console reflects that ambition. From one-click deployments with automatic TLS and custom domains, to real-time logs and monitoring that surface what actually matters, every detail was shaped by how modern teams work in practice. Database provisioning for PostgreSQL, MySQL, Redis, and MongoDB is built in. So is a command-line interface for engineers who prefer their terminal, project-scoped access keys for CI/CD pipelines, and role-based permissions for teams that need structure without friction. Everything runs on Skytells infrastructure with edge routing, automatic scaling, and built-in observability. There is no vendor lock-in. Bring your own API keys or use Skytells-managed quota. The choice is always yours. This is the beginning. The Console will continue to evolve as we work alongside the teams who use it, and we are committed to building the kind of platform we wish existed when we started this journey ourselves. Get started at console.skytells.ai or install the CLI with npm i -g @skytells/cli.