AtlasChat
Governed AI prompt control — deliberate, visible, and user-first.
AtlasChat is a governance-first browser extension that gives users explicit control over how prompts are sent to AI systems. No silent execution. No hidden state. No backend services.
Trust… But Verify.
What AtlasChat Does
AtlasChat operates at the UI boundary and governs how prompts are handled before they are sent.
Prompt Modes
Guard Mode (PME) Clarifies and structures user intent without executing requests.
Direct Mode (PMD) Sends prompts exactly as written — zero mutation.
Governance by Design
Mode and domain are always visible
No silent behavior
User intent always wins (fail-open safety model)
What AtlasChat Is Not
❌ Not an AI model
❌ Not a chatbot competitor
❌ Not a backend service
❌ Not a data store
❌ Not a compliance certification
AtlasChat governs user behavior, not vendor infrastructure.
Privacy & Architecture
Browser extension (Manifest v3)
No backend services
No external libraries
No prompt, response, or conversation storage
Metadata only (local, minimal)
Certification Status
Version: v1.0.0
Status: Certified & Frozen
Baseline: v1.0 Working Baseline
Checklist: Phases 1–4 complete
Known Issues: Cosmetic only (documented)
This release is immutable. All future changes will occur in v1.1+ following a new certification cycle.
Installation (Developer / Tester)
Download the release ZIP from Releases
Unzip locally
Chrome / Edge → Extensions → Enable Developer Mode
Click Load unpacked
Select the extracted folder
License
Personal, non-commercial use only. See LICENSE.txt for full terms.
Ethos
Trust… But Verify.
GhostShell Labs
Building tools that respect users, systems, and reality.