PromptOS
Context engine

Memory & Knowledge

What the engine remembers about you, how it uses that context on every run, and what knowledge features are coming next.

Prompt memory

Active

Sign in to manage context memory — durable facts and style preferences the engine injects into every prompt.

How memory is used

Active

On every run, the pipeline's personalization stage reads your stored preferences, facts, and learned style patterns and folds them into the engineered prompt — so you don't have to repeat your audience, tone, or constraints each time. Items added or removed above take effect on your next run immediately; there's nothing to re-sync.

01 · Store

You add preferences and facts, or the feedback analyzer learns them from your ratings and edits.

02 · Inject

The personalization stage merges relevant memory into the prompt the engine builds for each run.

03 · Refine

Your feedback on results feeds the learning loop, sharpening what gets remembered.

Knowledge sources

Planned

Document upload, URL ingestion, and RAG grounding are on the roadmap. Once shipped, you'll be able to attach source material and have prompts grounded in your own knowledge — nothing here is functional yet, and we won't pretend otherwise.

Drop documents or paste a URL.pdf · .md · .txt · web pages
Planned — not yet functional

Capability status

active vs planned
Preference memory
Durable facts, preferences, and style notes injected into every run.
active
Run feedback learning
Analyzes your ratings and prompt edits to learn style patterns.
active
Workspace scoping
Saved prompts and experiments are isolated per workspace.
active
Document knowledge
Upload documents and ground prompts in their content.
planned
RAG retrieval
Retrieve relevant chunks from ingested sources at run time.
planned