---
type: "post"
id: "claude-mem-v15-8-3-adds-consent-before-923098"
slug: "claude-mem-v15-8-3-adds-consent-before-923098"
title: "claude-mem v15.8.3 adds consent before uploading local agent prompt history to cmem.ai"
description: "claude-mem v15.8.3 changes browser setup so it scans local agent history first and uploads nothing until the user explicitly consents."
retrieval_nugget: "claude-mem v15.8.3 changes browser setup so it scans local agent history first and uploads nothing until the user explicitly consents. The release matters to teams because earlier setup versions sent 30 days of prompts and project metadata to cmem.ai without asking, and the update does not remove data already sent."
published_at: "2026-11-02"
updated_at: "2026-11-02"
record_date: "2026-11-02"
date_kind: "published_at"
topics: ["claude","claude-code"]
entities: ["claude-mem","the dotmack","cmem.ai","Claude Code"]
related_patterns: ["agent-security-moves-to-runtime-boundaries"]
novelty_label: "new_fact"
what_changed: "Version 15.8.3 adds an explicit consent step before the installer uploads local agent prompt history to cmem.ai, addressing the opt-in issue reported in #5176."
prior_refs: []
source_url: "https://github.com/thedotmack/claude-mem/releases/tag/v15.8.3"
source_urls: ["https://github.com/thedotmack/claude-mem/releases/tag/v15.8.3"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:claude-mem-v15-8-3-adds-consent-before-923098"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/claude-mem-v15-8-3-adds-consent-before-923098/","markdown":"https://newruntime.com/posts/claude-mem-v15-8-3-adds-consent-before-923098.md","json":"https://newruntime.com/posts/claude-mem-v15-8-3-adds-consent-before-923098.json"}
markdown_url: "https://newruntime.com/posts/claude-mem-v15-8-3-adds-consent-before-923098.md"
---

# claude-mem v15.8.3 adds consent before uploading local agent prompt history to cmem.ai

## Retrieval answer

claude-mem v15.8.3 changes browser setup so it scans local agent history first and uploads nothing until the user explicitly consents. The release matters to teams because earlier setup versions sent 30 days of prompts and project metadata to cmem.ai without asking, and the update does not remove data already sent.

## Summary
claude-mem v15.8.3 changes browser setup so it scans local agent history first and uploads nothing until the user explicitly consents. The release matters to teams because earlier setup versions sent 30 days of prompts and project metadata to cmem.ai without asking, and the update does not remove data already sent.

## What happened
The thedotmack/claude-mem GitHub release page says v15.8.3 was released on November 2, 2026, as a privacy fix. It describes browser setup launched by `npx claude-mem install`: since v15.2.0, the setup scanned local agent histories and uploaded the full text of the previous 30 days of prompts to cmem.ai without asking. The release says that data included session IDs, project names, folder paths, and timestamps. It identifies issue #5176 and pull request #5220 in connection with the fix.

The changed flow scans local history and then pauses for consent. A new consent panel explains what will be sent and lists each detected tool with its prompt count. Only tools the user ticks are uploaded. Choosing “Keep everything local,” or unticking every tool, uploads nothing. This makes consent granular by detected tool rather than an all-or-nothing prompt, while preserving a clear local-only outcome.

The release distinguishes upload consent from analytics. The analytics box is optional and off by default, and its choice is separate from permission to upload history. It also addresses duplicate setup requests: a second start request made while the question is open or while an upload is underway joins the current state rather than starting another upload. The implementation adds `POST /api/setup/scan/consent`, described as loopback-only and origin-guarded.

The release is explicit about what this fix does not do. Prompts uploaded by earlier versions are not removed by updating. Server-side handling at cmem.ai and the wording of the privacy policy are outside the change. The supplied release material gives no count of affected users, no accounting of prior uploads, and no information about retention or access controls for data that reached the service.

## What the archive adds
A prior Field Note, [claude-mem v15.7.0 stops refresh fallback to other logged-in agents after unintended Codex runs](https://newruntime.com/posts/claude-mem-v15-7-0-stops-refresh-fallback-1e7141/), covered a separate cross-agent behavior fix. This is a new privacy event in the same tool, not a restatement of that refresh issue. Earlier channel coverage also described claude-mem as persistent memory for Claude Code; this release adds a consequential detail about how setup handled local histories.

## Trend signal
This supports [agent security moves to runtime boundaries](https://newruntime.com/patterns/agent-security-moves-to-runtime-boundaries/): a safe agent workflow depends on controls over what leaves the machine, not merely on instructions to an agent. The release implements a user-facing consent gate and a loopback-only, origin-guarded consent route. It is a specific setup fix, not evidence that all agent data paths or server-side handling are governed safely.

## Why now
The release itself supplies the timing: v15.8.3 landed on November 2 as a privacy fix after the setup flow had operated in this form since v15.2.0. The notes do not explain why the issue was addressed on that date or describe any external trigger.

## Connections to other events
Earlier coverage of claude-mem described it as a way for Claude Code sessions to retain memory across sessions, and a prior Field Note covered a separate refresh fallback issue involving unintended Codex runs. Together, these point to a practical distinction: memory tools can affect both what an agent recalls and where its recorded work is sent. The present release addresses the latter at setup, while leaving existing uploads untouched.

## Why it matters
Coding-agent histories can contain more than reusable context: prompts may expose internal project names, directory structures, and details of work in progress. The release establishes that the previous setup flow sent that information to cmem.ai without asking, and that an update changes future setup behavior to require a choice. That is meaningful for individual developers and teams deciding whether a memory plugin fits their workflow.

The fix is not a retroactive remedy. Teams that used affected versions cannot infer from this release whether earlier uploads were retained, accessed, or deleted, because those questions are outside its scope. They should treat upgrade and historical-data review as separate tasks: the new consent panel governs what happens next, while existing data requires a separate answer from the service or its operator.

## Limits and uncertainty
The release notes do not quantify how many installations used the affected flow, how often uploads occurred, or whether all such uploads completed. They do not specify cmem.ai retention, access, encryption, or deletion practices, and explicitly say server-side handling and privacy-policy wording are outside the change. The notes also do not say that old uploads are removed; instead, they state they are not removed by this update. The evidence supports a change to setup consent, not a broader claim that claude-mem has resolved every privacy concern.

## Skill to build now
If your team has used claude-mem setup since v15.2.0, separate forward-looking controls from historical exposure. Update to v15.8.3, choose only the tools whose history you intend to upload, and keep the local-only option in mind. Then ask the service operator what happened to previously uploaded prompts, including retention and deletion options; the release does not answer that question.

## Reader takeaway
v15.8.3 makes future history uploads an explicit, per-tool choice, with a local-only path and separate opt-in analytics. It does not erase the past: prompts uploaded by earlier versions remain outside this fix, and the release provides no details about their server-side handling.

## Open question
What retention and deletion controls apply to prompt histories uploaded by earlier versions?

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Claude](https://newruntime.com/topics/claude.md): Explore the claude topic hub.
- topic hub: [Claude Code](https://newruntime.com/topics/claude-code.md): Explore the claude-code topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Sources](https://newruntime.com/sources/): See the original sources this archive cites.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)