Start a project
Choose New app, Import project, or Continue project. The active-project bar confirms which workspace future requests will use.
Ask anything, research the live web, work with files, or move useful outputs into a Project.
One intelligence workspace combines conversations, live research, first-party network data, persistent projects, Project Copilot, frontier reasoning, safe actions, workflows, memory, and measurable evaluations.
SkyMind profiles each task, chooses the minimum intelligence mode required, measures real model telemetry, and verifies important work before claiming completion.
See which mode, context, tools, verification, and token budget SkyMind would use before execution.
Higher modes add planning, tools, verification, and repair—not merely a larger completion budget.
Personal, conversation, project, network, and experience memory stay separate and are retrieved only when relevant.
Model scorecards and evaluation suites use observed telemetry and curated tests instead of invented benchmark claims.
SkyMind inspects the real project, explains the recommended update, prepares exact file changes, and waits for your approval before saving them.
Open, preview, activate, or delete a project from one permanent list.
Open a project to start a project-scoped conversation. Delete removes it from SkyMind after confirmation, and deleting the active project also detaches its context.
Checking GitHub connection…
SkyMind will discover the relevant real files automatically. Nothing is saved until you approve a proposed edit.
For example: “Modernize the admin area but preserve authentication,” or “Find why the mobile menu breaks and fix it.” SkyMind will inspect, explain, propose, and ask before writing.
Capacity, provider health, first-party route certification, incidents, and notifications are shown here instead of being hidden behind a separate experimental SkyMind.
Personal preferences, project instructions, private permissions, conversation context, network state, and experience signals remain separate rather than becoming one uncontrolled prompt.
Private tools stay off individually until permission is enabled.
Observed model reliability, latency, cost, evidence strength, and capability evaluations live here. SkyMind uses these measurements to improve routing instead of relying on model branding.
Read-only actions can complete immediately. Reversible and consequential actions wait here for approval.
Scheduled workflows run read-only steps automatically and pause before any write action.
Completed, failed, cancelled, expired, and undone actions are recorded with exact inputs and results.
Services are scored from real route history and quarantined after repeated failures.
Run checks, save baselines, and prepare staging packages directly through the Action Network.
The guide explains project-scoped coding, @file mentions, real file receipts, diffs, undo, conversation-to-project imports, actions, workflows, permissions, releases, and troubleshooting in plain language.
Choose New app, Import project, or Continue project. The active-project bar confirms which workspace future requests will use.
After creation or import, SkyMind shows the file receipt, opens the file tree, and displays the saved preview. Suggestions are never labeled as saved changes.
Open Project Copilot inside the workspace. Discuss, plan, edit, debug, or test while SkyMind stays attached to the active project and selected files.
Run quality checks, save a baseline, review changed files, and prepare a staging ZIP through the Action Network.