ClariLayer: carry selected work decisions across AI sessions

ClariLayer keeps selected facts, decisions and working rules available across AI projects and sessions. The idea is to preserve a choice you have already made, such as opening a launch page with a product demo, rather than explaining it again when a new conversation starts. Each saved item retains its source and the context where it applies.

Connect a client, then choose what to remember

The ClariLayer homepage lists guided connections for Claude Code, Cursor, Codex and claude.ai, along with compatible remote MCP clients. The general remote-client route requires Streamable HTTP MCP support. Setup includes checking the connection and saving a real decision from your work.

ClariLayer says connecting a client alone does not read conversation history or install ongoing capture. History import is a separate choice: select a supported local source, choose a space and extraction provider, preview the batch, then accept the useful context. Ongoing capture can be enabled for selected sources.

Client connectivity and history extraction are separate capabilities. The captured page says its qualified extraction uses a version-specific Claude CLI profile; OpenAI and Cursor extraction providers are unavailable, and native Cursor history qualification remains limited. A listed client connection should not be read as support for importing that client's complete local history.

Keep a decision within its proper scope

A memory can apply to a project or an authorized space. ClariLayer says separate spaces do not merge automatically and a project-specific decision does not become a universal rule. You can inspect a saved item's source, correct it while retaining a reference to an earlier version, or forget it within the relevant scope.

For example, a definition of an active customer may belong in a revenue-analysis space, while a launch-page choice belongs to one launch project. Review that scope during import so an old preference does not become an instruction for unrelated work. The publisher also states that general work memory is user-controlled, not independently verified or guaranteed to appear in every answer.

Understand extraction and evaluation limits

ClariLayer says it stores extracted context rather than a raw transcript archive. Selected source content still passes through the chosen extraction provider, and optional semantic processing has separate consent and provider disclosure. Check those terms before supplying sensitive conversations.

The page presents an internal paired analytics evaluation with results for thirty-eight questions. It labels this as the company's evaluation, not an independent benchmark or a guarantee across tasks. This overview does not establish a general accuracy improvement from saving memory.

The current personal offer starts free without a credit card, but history processing can consume the chosen extraction provider's allowance or incur its usage costs. Begin with one supported source and a small set of decisions you can inspect; confirm that the next session receives the intended context before relying on it for wider work.