What Atlas Neural Is
Neural is the cyberpunk front end for Atlas Omni-Core. The animated brain represents live regions such as the core router, agent mesh, memory vault, skills, automation, research, security, voice, and shop systems. It is not a separate assistant: commands entered here travel through Atlas and use the same backend, permissions, services, and stored knowledge as the normal Atlas interface.
What Atlas Can Do
- Answer questions and research information.
- Inspect, create, edit, test, and organize software.
- Run approved server tasks, containers, and automation.
- Monitor services, memory, agents, queues, and system health.
- Create local CPU-generated images and Etsy product packs.
- Manage shop goals, listings, publishing workflows, and profit signals.
- Work through Neural, the normal Atlas UI, voice, phones, or WhatsApp.
How Tasks Move
Your request first reaches Atlas Core, which identifies the goal and domain. The skill retriever selects a small set of relevant procedures. The orchestrator routes the task to a suitable specialist agent. Atlas then executes available tools, records status, checks the result, and returns the outcome. Simple work can stay with one agent; larger work can be divided between specialists.
Agents and Concurrency
Agents are workers, while skills are instructions and capabilities they can use. Atlas dynamically gives an agent the skills relevant to its current task instead of permanently limiting one agent to one skill. This server is currently configured for up to two specialist tasks at the same time, protecting CPU and memory while still allowing parallel work. Atlas can queue additional jobs and start them as capacity becomes available.
How Atlas Memory Works
Working memory holds the current conversation and recent task context. Persistent memory lives in Atlas data, vault files, project records, and saved preferences so useful knowledge can survive restarts. Voice-profile memory can keep information associated with an enrolled person. WhatsApp also keeps limited recent chat history so follow-up requests make sense.
Memory is retrieved only when relevant; Atlas does not place every saved item into every prompt. Skills are not memories: they are reusable operating procedures. Deleting a chat does not automatically delete independently saved vault or profile memory.
Skills and Learning
Skills describe how Atlas should perform specialized work, which tools are allowed, required checks, and risk boundaries. Atlas searches the skill registry by name and relevance, then assembles the best skill context for the task. Lessons from failures can be written into durable guidance, such as the WhatsApp recovery lessons, so the same infrastructure error is handled better next time.
Local Server Execution
Atlas runs its control plane and connected services on your server. CPU-only generation is available for digital products and graphics. Docker isolates major services, while the API provides health, agent, shop, phone, voice, and automation controls. Some AI reasoning may use configured external model services; “local” describes where Atlas executes and stores its server work, not a promise that every model request is offline.
Voice, Research and Communication
The Speak tab opens Atlas voice presence inside Neural: microphone conversation, voice recognition, local speech generation, and spoken replies. Voice can retrieve current time and date, Open-Meteo weather, Google News headlines, and read-only web evidence for fact-checks. Live answers display source links and Atlas must state uncertainty when evidence is weak or conflicting. Phone and WhatsApp bridges provide additional ways to send work and receive results.
Automation and Self-Healing
Atlas can schedule durable work, route jobs, retry recoverable failures, and monitor task state. Self-healing is bounded: Atlas can apply known safe recovery steps, but authentication failures, destructive actions, payments, publishing, or uncertain changes may still require your approval. The live brain speeds up and brightens when agents, tasks, or server load indicate active processing.
Security and Control
Authentication protects Atlas interfaces. Sensitive credentials should remain in server secrets rather than UI code or logs. Read-only checks can run safely, while writes and high-risk operations should follow confirmation rules. Atlas records operational state and exposes what is running, but it cannot guarantee that third-party services, network links, or user-provided files are safe without inspection.
Interfaces and Regions
Core routes intent. Agent Mesh delegates work. Memory Vault retrieves context. Skill Lattice supplies procedures. Automation manages durable jobs. Research gathers evidence. Security watches boundaries. Voice handles speech and people. Shop Engine manages Etsy workflows. Select a region in the brain or Systems panel to inspect it.
What Atlas Does Not Do Automatically
Atlas does not bypass passwords, invent missing API access, guarantee profit, or make every risky decision without you. It cannot complete external account setup until valid credentials and permissions exist. Queued work is not the same as completed work, and generated output should be reviewed before publishing or production use. These limits keep the system understandable and under your control.