One API: save, find, context. The exact words, for every turn.
Your app sends Remiq what it sees: messages, documents, events and code. Prism, our model, keeps the words with who said them, when, and who may read them. Before each turn it hands your model the few lines that matter, each with an id to cite.
What Remiq keeps
| Kind | What it is | Example |
|---|---|---|
| Records | The original words, with who, when, source and who may read | "We moved the launch to 14 April." |
| Graph | People, projects, decisions and dates, linked to the records that mention them | Priya → Launch → 14 April |
| Values | The current value of anything that changes, and the ones it replaced | launch: 14 April (was 31 March) |
One turn, step by step
- 01Your app asks for context with the user's new message.
- 02Prism, our model, finds the lines that matter: search, the graph and current values, only what this person may read.
- 03Your model reads them, answers and cites the ids, like [M41].
- 04The turn's new words are saved. Up to 500 tokens come with the same credit.
Privacy and deletes
Each person sees only what they may see, and a hidden record is never hinted at. A delete removes the record from every index and cache. Retries never charge twice.
Use it from your code
from prism_client import Prism prism = Prism(api_key="prism_...", base_url="https://remiqlabs.com") prism.add("We moved the launch to 14 April.", user="priya") ctx = prism.context("When is the launch?", user="priya")
Python and TypeScript clients, plain HTTPS, and an MCP server for Claude, ChatGPT, Cursor, Claude Code and any MCP client.