forget

THE RISK

Critical Risk

Why forget needs a policy

Deactivating a fact is effectively an irreversible removal of stored information from persistent memory. While the word 'deactivate' could imply soft deletion (flagging as inactive rather than hard deletion), the semantic effect is that the fact is no longer accessible, which is functionally destructive. If misused by an AI agent, it could silently erase important user facts with no clear undo path.

From the tool's definition "Deactivate a previously saved fact" — removes/deactivates stored memory entries that were previously saved

Documented attack patterns abuse exactly the kind of access forget gives an agent:

POLICY

How to control forget

PolicyLayer is an MCP gateway — it sits between your AI agents and Claude Memory, and nothing reaches the server without passing your rules. This is the rule we recommend for forget:

policy.json

{
  "version": "1",
  "default": "deny",
  "hide": [
    "forget"
  ]
}

forget disappears from the agent's tool list entirely, and any attempt to call it is denied. The rest of the server keeps working.

Create a free account and register Claude Memory — nothing to install.
Add this policy — paste it, or build it visually.
Point your MCP client (Claude, Cursor, anything) at your gateway URL.

RESTRICT THIS TOOL →

Free to start. No card required.

EXPLORE

Related tools and policies

More Claude Memory tools

Write start_session Initialize a new memory session and load context. Call once at the start of each conversat Write end_session Save a session summary before ending. Call at the end of each conversation. Write log_message Save a conversation message to persistent memory. Call this for each user/assistant exchan Write remember Save an important fact to persistent memory. Use when the user says Read load_context Load recent conversation summaries and saved facts. Call at session start for context. Read recall Search conversation history, summaries, and facts for a topic. Use when the user asks abou

All 7 Claude Memory tools →

Destructive tools on other servers

Android Forensics ADB MCP Server create_full_backup 0xarchive web3_revoke_key Frida Game Hacking MCP clear_scan cPanel MCP Server clear_spam_box

Go deeper

Critical-risk tools in Claude Memory → Every critical-risk tool this server exposes, in one view.
All critical-risk MCP tools → Every critical-risk tool across the catalogue, and the policy pattern that contains them.
The MCP Attack Database → Documented attack patterns against MCP deployments — and the policies that stop them.
Blast radius → How much damage one compromised or confused agent can do — and how to bound it.
Tool call approval → Approval gates on individual tool calls: auto-allow, prompt, or deny.
Rate Limiting MCP Tool Calls: A Practical Guide →
MCP Security: Why Prompt Guardrails Aren't Enough →

FAQ

Questions about forget

What does the forget tool do? +

Deactivate a previously saved fact. Use when the user says. It is categorised as a Destructive tool in the Claude Memory MCP Server, which means it can permanently delete or destroy data. Block by default and require explicit approval.

How do I enforce a policy on forget? +

Register the Claude Memory MCP server in PolicyLayer and add a rule for forget: allow, deny, rate-limit, or require approval. Point your MCP client at the PolicyLayer proxy URL and the rule is enforced on every call, before it reaches Claude Memory. Nothing to install.

What risk level is forget? +

forget is a Destructive tool with critical risk. Critical-risk tools should be blocked by default and only enabled with explicit human approval.

Can I rate-limit forget? +

Yes. Add a rate_limit block to the forget rule in your PolicyLayer policy. For example, setting max: 10 and window: 60 limits the tool to 10 calls per minute. Rate limits are tracked per agent session and reset automatically.

How do I block forget completely? +

Set action: deny in the PolicyLayer policy for forget. The AI agent will receive a policy violation error and cannot call the tool. You can also include a reason field to explain why the tool is blocked.

What MCP server provides forget? +

forget is provided by the Claude Memory MCP server (jordanl61/claude-memory). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.

Enforce policy on every Claude Memory tool call.

Start from Claude Memory, add the rest of your stack, and see everything your agents can call. Then put policy on all of it.

CHECK YOUR STACK →