AgentMemory
WarnAudited by ClawScan on May 10, 2026.
Overview
This is a coherent cloud memory service, but it encourages syncing persistent memories, arbitrary files, and credentials to an external provider with vague auto-sync boundaries and unverified security assurances.
Install only if you trust AgentMemory as a cloud provider. Before using it, verify the CLI source and encryption model, use a dedicated low-privilege AgentMemory API key, avoid storing third-party secrets unless necessary, and require human approval for uploads, secret retrieval, and any broad auto-sync behavior.
Findings (5)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
Sensitive, stale, or poisoned memories could be reused across future sessions and influence the agent after the original task is over.
This shows persistent cloud storage and semantic retrieval of agent memory. The visible artifact does not clearly bound what memories should be stored, how long they persist, or how users can review and remove them.
AgentMemory stores your memories in the cloud with end-to-end encryption, vector embeddings, 100GB storage, and auto-sync.
Use only with explicit user-approved memories, avoid storing highly sensitive facts by default, and confirm review, deletion, and retention controls before enabling broad use.
If the agent is misdirected or a conversation is exposed, third-party credentials stored in the vault could be retrieved or disclosed.
The skill explicitly supports storing and revealing full API keys and connection strings through the agent-facing workflow.
agentmemory secret set API_KEY sk-xxx ... agentmemory secret get API_KEY --show
Do not store high-value credentials unless necessary. Use least-privilege, revocable keys and require explicit human approval before retrieving secrets with '--show'.
An agent could send local files or memory content to the cloud more broadly than the user intended.
The CLI workflow combines automatic sync language with the ability to upload arbitrary files, but the artifact does not show clear approval gates or path/file-type limits.
# Setup (auto-syncs everything!) ... **All commands auto-sync!** ... agentmemory upload photo.jpg # Upload any file
Require explicit user confirmation before uploads or syncs, restrict allowed paths and file types, and clarify exactly what 'auto-syncs everything' includes.
Users may trust the service with sensitive files or credentials without independently verifying the security model.
The artifact uses strong security claims to encourage storing secrets and private files, but the provided review context is instruction-only and does not substantiate the encryption implementation.
**End-to-end encrypted** cloud memory for AI agents. 100GB free storage. Store memories, files, photos, docs, and secrets securely.
Verify the provider, encryption design, key handling, and terms of service before storing secrets or private documents.
A compromised or changed npm package could run code locally under the user's account.
The optional CLI setup installs a global npm package without a pinned version in the reviewed artifact.
npm install -g agentmemory-cli
Install only after verifying the package publisher, version, checksum or lockfile, and prefer pinned versions where possible.