Memory with receipts.
Remiq keeps what your users and agents say, and gives your model the exact context for every turn: the original words, dated and sourced. Your model, your keys.
Save. Find. Context for every turn.
Save
Every message, document and event your app sees. Saved once, with who said it, when, and who may read it.
Find
People, projects, decisions and dates. The current value of anything that changes, and its history.
Context
The exact words your model should read for this turn, each with an id it can cite. Your model answers.
Prism, our own model, does the work inside Remiq: it keeps the original words, links the people and projects in them, versions every change, and finds the few lines that matter for each turn.
12 Mar · Priya: We moved the launch to 14 April. The budget stays at ₹40 lakh.
What makes it different
The original words
Remiq never rewrites what people said. Your model reads the source, not a summary of it.
Changes, tracked
A value moves twice: Remiq knows the current one and shows the two it replaced.
Privacy per person
Each person sees only what they may see. A hidden record is never hinted at.
Safe by default
Retries never charge twice. A delete removes the record from every index and cache.
| Question | An AI-extracted memory | Remiq |
|---|---|---|
| What is kept | A model's rewrite of what was said | The original words |
| Proof for each line | None | Date, source and an id to cite |
| When a value changes | Overwritten | The current value and the ones it replaced |
| The model doing the work | A general chat model | Prism, our own model, built for memory |
Three products, one memory.
Remiq
The memory API and the Console: save, find, context, with evidence for every line.
Read the docs →Glass
Select anything in any app and save it to your memory. Context comes back when it matters.
After launchKeeper
Stops prompt attacks hidden inside memories before they reach your model.
After launch