AI agents use memstem_upsert to create or update resources in Memstem — usually the action step of a workflow, after the agent has gathered context. Every call changes real data in your Memstem environment.
Upsert operations create or modify data in a memory store. This is reversible (data can be corrected, deleted, or overwritten later), so it falls under Write rather than Destructive. The severity is medium because misuse could corrupt shared memory state affecting multiple AI agents, but the blast radius is contained to memory/data rather than external systems or finances.
From the tool's definition Tool name 'memstem_upsert' indicates an update or insert operation. Upsert semantics combine create and modify operations, which are Write category actions. The empty description limits precision but the name strongly suggests a reversible data modification.
Attacks that exploit this kind of access
memstem_upsert. It is categorised as a Write tool in the Memstem MCP Server, which means it can create or modify data. Consider rate limits to prevent runaway writes.
Register the Memstem MCP server in PolicyLayer and add a rule for memstem_upsert: 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 Memstem. Nothing to install.
memstem_upsert is a Write tool with medium risk. Write tools should be rate-limited to prevent accidental bulk modifications.
Yes. Add a rate_limit block to the memstem_upsert 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.
Set action: deny in the PolicyLayer policy for memstem_upsert. 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.
memstem_upsert is provided by the Memstem MCP server (memstem/memstem). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.
Every MCP server has a record like this.
Type a name, get the same breakdown: verified identity, auth posture, risk grade, capabilities, recommended policy.
Teams ship this data inside their own products. See what a licence covers →