Introducing the Skytells CLI — Ship, Scale, and Observe from Your Terminal

Introducing the Skytells CLI — Ship, Scale, and Observe from Your Terminal

April 13, 2026 — Today we're releasing the Skytells CLI as a generally available tool. It's an open-source command-line interface that gives engineering teams direct terminal access to the full Skytells platform — from project scaffolding and app deployment to GPU instance management, orchestration workflows, and production observability through Cognition.

npm install -g @skytells/cli

One install. One skytells command. Everything your team needs to build, deploy, and operate — without switching to a browser.

The Problem We Kept Running Into

Over the past year, we've expanded the Skytells platform considerably. We launched globally distributed GPU infrastructure with NVIDIA H100 and A100 availability, built Orchestrator for workflow automation, introduced Cognition for full-stack observability, and added edge computing across new regions. The Skytells Console covers all of this, and it works well for visual management.

But there's a pattern we noticed internally first, and then heard repeatedly from customers: the engineers doing the actual deployment and operations work were losing time to context switching.

A typical deployment workflow looked like this: write code in the editor, switch to the console to trigger a deploy, open another tab to check logs, switch again to update an environment variable, go back to check the deployment status. Each switch is small, but the cumulative cost — measured in focus, not minutes — is significant. There's a body of research on this. Gloria Mark's work at UC Irvine showed it takes over 23 minutes on average to return to deep focus after an interruption. Even brief app switches contribute to that fragmentation.

We didn't build the CLI because a command-line tool is fashionable. We built it because every platform operation our customers perform should be accessible in the same environment where they already work: the terminal.

What the Skytells CLI Covers

The CLI is not a subset or a companion. It covers the full operational surface of the platform — every action available in the console is available through the CLI. Here's what that looks like in practice.

Secure Authentication with Scope Control

Authentication matters more in a CLI context because credentials live on disk. We designed authentication with that reality front and center.

The recommended path uses an OAuth device flow. You run skytells login, a browser window opens, you approve the request, and the CLI receives a scoped token. No credentials are pasted into the terminal, no secrets are echoed to stdout.

# Secure browser-based device flow
skytells login

# Scope your session to specific permissions
skytells login --scopes inference,projects.read,deployments.read

For headless environments — CI runners, Docker containers, remote servers — personal access tokens are supported. These can be generated at console.skytells.ai/settings/tokens with specific scopes: inference, projects.read, projects.write, deployments.read, deployments.write, keys.read, keys.write, and account.read.

# For CI/CD pipelines and headless environments
skytells login --token

Every credential file is written with 0600 permissions. Token storage uses ~/.config/skytells/credentials.json — a standard, inspectable path. No opaque keychains, no hidden state.

Project-scoped access keys add another layer of control. These keys (prefixed sk_proj_) restrict CLI operations to a single project, so a team member or pipeline with a project key cannot interact with resources outside that project boundary:

skytells link sk_proj_abc123...
skytells whoami --json

Full Project and Application Lifecycle

Managing projects and applications from the terminal means going beyond listing and creating. The CLI supports the entire lifecycle: creating projects, adding and configuring apps, inspecting runtime state, updating settings, starting, stopping, restarting, redeploying, and removing resources — each with proper confirmation guards that must be explicitly bypassed with --force.

# Create a new project and add an application
skytells projects add production-backend
skytells apps add api-service --type web

# Inspect the running state of an application
skytells apps inspect abc123 --json

# Restart without switching to the console
skytells apps restart abc123

# Full redeployment
skytells apps redeploy abc123

For teams managing multiple applications across projects, the ability to script these operations — and combine them with --json output for parsing — replaces dozens of manual console interactions per week.

Deployment Pipeline Integration

Deploying from the CLI is straightforward. But the design goes further than skytells deploy. The deployment subsystem supports listing deployment history with filtering by app, paginating through results, and outputting structured JSON — which makes it practical to build deployment dashboards, audit trails, and rollback scripts on top of the CLI.

# Deploy an application
skytells deploy api-service

# Review the deployment history for a specific app
skytells deployments ls --app abc123 --limit 10 --json

In a typical CI/CD pipeline, a deployment step becomes a single line:

# Example: GitHub Actions step
- run: skytells deploy api-service --json
  env:
    SKYTELLS_TOKEN: ${{ secrets.SKYTELLS_TOKEN }}
    SKYTELLS_ACCESS_KEY: ${{ secrets.SKYTELLS_ACCESS_KEY }}

No custom API clients. No shell scripting around curl. The CLI handles authentication, request formatting, error handling, and output.

Database Provisioning and Management

The Skytells platform supports managed databases across five engines: PostgreSQL, MySQL, MariaDB, MongoDB, and Redis. The CLI exposes full control over database provisioning, configuration, backup management, and lifecycle operations.

# Provision a PostgreSQL database with custom configuration
skytells databases add analytics-db postgres \
  --db-name analytics \
  --db-user app_service \
  --description "Analytics pipeline storage"

# Enable automated backups with a custom schedule
skytells databases set db-id backup_enabled true
skytells databases set db-id backup_schedule "0 2 * * *"

# Configure external access for development
skytells databases set db-id external_port 5432

Database lifecycle commands — start, stop, deploy, rm — follow the same pattern as app management, with --force flags for unattended scripts and --json for programmatic output.

Environment Variable Management at Scale

Environment variables are one of the most frequent operations teams perform. The CLI supports setting multiple key-value pairs in a single command, scoped to a project or a specific application:

# Set multiple variables at once — project-wide
skytells env set DATABASE_URL=postgres://... REDIS_URL=redis://... NODE_ENV=production

# Set variables scoped to a specific app
skytells env set --app api-service SECRET_KEY=sk_live_... LOG_LEVEL=info

# Audit current variables
skytells env ls --app api-service --json

This is particularly valuable during incident response, where updating a configuration value across services needs to happen in seconds — not through a multi-step console workflow.

Custom Domains and Team Management

Domain management and team collaboration are available directly through the CLI:

# Add a domain and associate it with an application
skytells domains add api.example.com --app api-service

# Review current team members
skytells members ls --json

Real-Time Log Streaming

Logs are where engineers spend a meaningful portion of their debugging time. The CLI supports streaming container and deployment logs in real time over server-sent events (SSE), with filtering by log type, deployment, and tail depth:

# Stream live logs from a running application
skytells logs api-service --follow

# View the last 100 lines of deployment logs
skytells logs api-service --type deployment --tail 100

# Stream logs from a specific deployment for post-mortem analysis
skytells logs api-service --deployment dep-abc123 --follow

The --follow flag keeps the connection open and streams new log entries as they arrive — no polling, no refresh. Combined with standard Unix tools like grep, awk, and jq, this turns the terminal into a live operations dashboard.

Cloud Infrastructure and GPU Computing

The Skytells cloud infrastructure spans multiple regions globally, with support for standard compute instances and high-performance GPU instances (NVIDIA H100, A100). The CLI provides full control over cloud instance lifecycle: provisioning, inspection, networking, and teardown.

# Deploy a standard compute instance
skytells cloud deploy \
  --region ewr \
  --plan vc2-1c-1gb \
  --os 387 \
  --label "staging-server" \
  --hostname "staging-01"

# Deploy an H100 GPU instance for model training
skytells cloud deploy \
  --region ewr \
  --plan gpu-h100-1 \
  --os 387 \
  --label "training-h100"

# Deploy an A100 GPU instance for inference workloads
skytells cloud deploy \
  --region lax \
  --plan gpu-a100-1 \
  --os 387 \
  --label "inference-a100"

Instance management supports VPC networking, firewall groups, IPv6 configuration, and tagging — all configurable at deploy time:

skytells cloud deploy \
  --region ewr \
  --plan gpu-h100-1 \
  --os 387 \
  --label "secure-training" \
  --vpc vpc-id \
  --firewall-group fw-id \
  --ipv6 \
  --tags "ml,training,production"

Lifecycle operations — start, halt, reboot, destroy — each have confirmation prompts by default. For scripted environments, --force bypasses them.

For teams running AI workloads, this means GPU instances can be provisioned, used for training or inference, and destroyed programmatically — reducing idle compute costs and giving engineers direct control over their infrastructure spend.

Orchestrator: Workflow Automation from the Terminal

Orchestrator handles complex, multi-step workflows — ETL pipelines, model retraining cycles, scheduled batch jobs, and event-driven processing chains. The CLI brings full visibility into this system:

# List all defined workflows
skytells workflows ls --json

# View orchestrator health and execution statistics
skytells orchestrator overview

# Review completed executions
skytells orchestrator executions --status completed --limit 20

# Inspect a specific execution with full step-by-step logs
skytells orchestrator inspect exec-abc123 --json

For operational planning, the CLI exposes usage and metrics data over configurable time windows:

# View execution metrics for the last month
skytells orchestrator metrics --from 2026-03-01 --to 2026-03-31

# Review resource usage
skytells orchestrator usage --from 2026-04-01 --to 2026-04-13

This data feeds directly into capacity planning decisions — understanding how many workflow executions run per day, what the average duration looks like, and where bottlenecks occur.

Cognition: Production Observability Without Leaving the Terminal

Cognition is the observability layer of the Skytells platform. It provides application health monitoring, error tracking, security event logging, anomaly detection, runtime performance snapshots, and real-time event streaming. The CLI surfaces all of this.

# High-level health overview for the last 24 hours
skytells cognition overview --project proj-id --hours 24

# List recent application errors
skytells cognition errors --limit 20 --json

# Review security events and alerts
skytells cognition security --project proj-id

# Detect anomalies across your applications
skytells cognition anomalies --project proj-id --limit 10

# View runtime performance snapshots
skytells cognition runtime --project proj-id

Two commands deserve particular attention.

Real-time event streaming uses the --since parameter to implement a polling pattern — useful for integrating with alerting tools or custom dashboards that consume structured JSON:

skytells cognition events --project proj-id --since last-event-id --json

Time-series metrics provide granular operational data over custom windows, suitable for reporting, SLA tracking, and capacity modeling:

skytells cognition timeseries --project proj-id --hours 48 --json

For on-call engineers, having cognition errors, cognition security, and logs --follow available in adjacent terminal panes turns an incident response workflow into something fast and self-contained.

Designed for Unattended Execution

Every decision in the CLI was made with automation in mind.

Structured output. Every command that returns data supports --json. This means CLI output can be piped directly to jq, consumed by scripts, or parsed by CI/CD tools — without scraping human-readable tables.

Environment variable overrides. Authentication can be handled entirely through environment variables, removing any interactive prompt from automated pipelines:

Confirmation guards. Every destructive action — deleting apps, destroying cloud instances, removing databases — requires explicit confirmation. In CI/CD, --force bypasses the prompt. This is a deliberate design choice: unattended pipelines should declare their intent clearly in the script, not rely on interactive defaults.

Exit codes. The CLI uses standard exit codes, so it integrates cleanly with shell conditionals, set -e scripts, and CI systems that rely on process return codes for pass/fail determination.

Security Architecture

We take credential management seriously, and the CLI reflects that. Every design decision in authentication and storage was reviewed against our Trust Center standards.

• Credential isolation. Tokens are stored in a single file at ~/.config/skytells/credentials.json, created with 0600 permissions — readable only by the file owner
• Scope-limited tokens. Personal access tokens support fine-grained scopes. A token granted projects.read and deployments.read cannot write or delete anything
• Project-scoped access keys. sk_proj_ keys restrict all operations to a single project boundary. A compromised project key cannot access other projects
• No credential echoing. The --token login flow uses a secure prompt. Tokens are never printed to stdout or logged
• Configurable storage. The SKYTELLS_CONFIG_DIR environment variable lets teams point credential storage to an encrypted volume or a secrets manager mount

This architecture means the CLI can be deployed across development machines, CI runners, and production automation environments with appropriate security boundaries at each layer.

Open Source, MIT Licensed

The Skytells CLI is released under the MIT License and the source code is available on GitHub. We believe that developer tools should be inspectable. Engineers should be able to read the code that manages their infrastructure, verify its behavior, and contribute improvements.

The repository includes a changelog and a security policy for responsible vulnerability reporting.

Getting Started

The CLI requires Node.js 18.0.0 or later and a Skytells account.

# Install globally
npm install -g @skytells/cli

# Verify installation
skytells --version

# Authenticate
skytells login

# Link to a project
skytells link sk_proj_your_access_key

# View your project and status
skytells project
skytells status

From here, you have the full platform at your fingertips. Deploy an app, stream logs, provision a database, spin up a GPU instance — all from the same terminal session.

Full documentation, including command references and integration guides, is available at Skytells Learn.

What Comes Next

Version 1.0.0 is the foundation. The decisions we made here — structured output, scope-based auth, full platform coverage — are intentional. They create a stable surface for what we're building next.

On our roadmap: interactive guided workflows for common operations, plugin architecture for third-party extensions, deeper integration with our inference APIs and edge computing layer, and skytells init for bootstrapping new projects from templates.

We will continue shipping updates through npm and the GitHub repository. Breaking changes, if any, will follow semantic versioning and be documented in the changelog.

About Skytells

Skytells is an AI research and infrastructure company building enterprise-grade platforms, intelligent systems, and globally distributed cloud infrastructure. The company serves customers across technology, healthcare, finance, and government sectors, offering managed AI services, GPU computing, and production observability. Learn more at skytells.ai.