Overview

PRIMA Memory gives Claude Code access to its own conversation history. Claude Code saves every conversation to a file on your computer, but a new session starts with no knowledge of the ones before it. PRIMA Memory reads those saved conversations and gives Claude a set of tools to search them, open them and summarise them, so you can pick up where you left off without explaining the work again.
This manual describes PRIMA Memory 2.7.0, and the PRIMA Memory release notes list the changes in each release.
What You Can Do With It
Section titled “What You Can Do With It”- Ask Claude what you worked on today, yesterday, last week or last month.
- Find the session where you made a decision and read the conversation that led to it.
- See every change Claude made to a file across your sessions, with the explanation Claude gave when it made the change.
- Resume an earlier session with the
/resumecommand. - Get a summary of a period of work, with the projects, files and topics it covered.
You do not call the tools yourself. You ask Claude in plain words, for example “what did we decide about the Spring Workshop venue last week?”, and Claude chooses the tool that answers the question.
How It Works in Brief
Section titled “How It Works in Brief”PRIMA Memory searches your history in two ways and combines the results by default. Keyword search reads the conversation files directly each time you search, so it always covers everything. Semantic search finds conversations by meaning, using an index that a small embedding model builds on your computer. A search for “venue” can then find a session where you only wrote “room booking”.
The index lives in a .prima-memory/ folder inside your workspace. The first time a semantic search runs, PRIMA Memory downloads the embedding model (about 23 MB) from Hugging Face. That download is the only network access, and no conversation content leaves your computer. The Privacy and Your Data page lists everything PRIMA Memory reads, writes and sends.
The Seven Tools
Section titled “The Seven Tools”| Tool | What it does |
|---|---|
search_history |
Searches your history by keyword, by meaning or both, with optional time and project filters |
get_recent_sessions |
Lists your most recent Claude Code sessions in this workspace |
get_session_details |
Opens one session and shows its conversation |
get_file_history |
Shows every recorded change to a file across sessions |
summarise_period |
Summarises the work done over a day, week or month |
rebuild_index |
Clears the semantic index and rebuilds it in the background |
get_index_status |
Reports how much of your history the semantic index covers |
The Tools Reference describes each tool in full.
What It Reads
Section titled “What It Reads”In a standard install PRIMA Memory reads the Claude Code conversations saved for the current workspace. It also reads three other sources when they exist on your computer, which are a shared project log inside the workspace, VS Code Copilot Chat sessions and Codex sessions. It only ever reads these files and never changes them. How Indexing Works lists each source and its scope.
Requirements
Section titled “Requirements”- AI Business OS, which installs PRIMA Memory from the email you receive when you buy it (see Installing PRIMA Memory).
- Claude Code, which creates the conversation files PRIMA Memory reads.
- Node.js 22.5 or later.
The developer builds and tests PRIMA Memory on macOS. It may run on Windows through WSL or Git Bash, but Windows has no direct support. The full list for every product is on the Requirements page.
How It Fits With the Other Products
Section titled “How It Fits With the Other Products”- AI Business OS installs PRIMA Memory and provides the
/saveand/resumecommands./saveworks without PRIMA Memory./resumewith a time, or with nothing after it, falls back to the session history in the project register, and/resumewith a project ID or a search word needs PRIMA Memory. - PRIMA Project Management uses PRIMA Memory in its
/resumecommand when the server is available, while its/daybriefing works without it (see Resuming a Session). - PRIMA CRM is a separate server, and neither one reads the other’s data. When
/saveor/nightreports the CRM records it logged, that report becomes part of the conversation, so a later session can find it withsearch_history.
How the Products Fit Together shows the whole suite.
Credits
Section titled “Credits”Parts of PRIMA Memory derive from claude-mems by Joe Tustin, which the PRIMA Memory licence file credits under the MIT licence.

