ALIENDISTRICT51 Agentic Command
A local-first, agentic operating layer for a person's own work: durable structured memory, scope-aware context, knowledge navigation, and permission-controlled tools that only reach outside the system when the user explicitly authorizes it.
What Agentic Command is
Agentic Command is a desktop application built around a local "Cognitive Core" — a structured memory and reasoning support system that runs on the user's own machine. Instead of storing notes as loose text, it organizes work into durable records: tasks, decisions, learnings, procedures, and artifacts, each tied to an explicit scope (a project, a client, or a workspace) so information never bleeds across contexts it doesn't belong to.
On top of that memory layer sits a set of agent-facing capabilities that let the system retrieve the right context, navigate curated knowledge, understand relationships between pieces of work, and — only when the user allows it — take bounded, permission-gated actions using external tools and services.
Core architectural concepts
Cognitive Core
A local, SQLite-backed service that stores operational memory and provenance. It runs entirely on the user's machine and is optional to the base application.
Durable memory
Tasks, decisions, learnings, procedures, and artifacts are recorded with timestamps and provenance, rather than existing only as transient conversation history.
Scoped context
Every record belongs to an explicit scope. Reads default to that exact scope, so information for one project or client is never silently mixed with another.
Knowledge navigation
A curated, link-based map over durable notes lets the system find relevant material by following authored relationships, not by guessing.
Relationship intelligence
A derived, rebuildable graph of relationships between code, documentation, and notes — used only as a supplementary lookup, never as a second source of truth.
Agents
Role-based views over the same underlying memory, each seeing only what its role and scope permit.
Skills
Versioned, inspectable definitions of a repeatable job. Published versions are immutable, so behavior can't silently change after the fact.
Capability composition
Higher-level capabilities are built by combining existing skills, not by duplicating logic.
Permission-controlled tools & connectors
Any action that reaches outside the Cognitive Core goes through an explicit permission ladder, with higher-impact actions requiring a specific, time-bounded approval.
Optional external service integrations
Agentic Command's core functionality does not require connecting any external account. A user may optionally authorize specific external services — such as Google services described below — to extend what the application can do on their behalf. Nothing external is contacted until the user explicitly connects and authorizes it.
Authentication note: Any Google connection uses Google's own OAuth authorization screen. Agentic Command never asks for, sees, or stores a Google account password.
Gmail (optional, user-authorized)
When a user explicitly connects Gmail, Agentic Command can, within that authorization:
- Retrieve a bounded set of messages relevant to a specific, user-requested task
- Search messages, and read threads and labels
- Prepare draft messages
- Send a message only after the user has explicitly approved that specific send
Google Calendar (optional, user-authorized)
When a user explicitly connects Google Calendar, Agentic Command can, within that authorization:
- Retrieve event details relevant to a requested time range
- Check free/busy availability
- Prepare a proposed event
- Create or update a calendar event only after the user has explicitly approved that specific action
Approved files and documents
Document access is restricted to specific local folders the user has explicitly approved in advance. Agentic Command does not have general, unrestricted access to the filesystem.
No Google account is connected by default. These capabilities describe what becomes available only after a user completes Google's own authorization flow and grants the corresponding permission.