AI agents use memory_manager to create or update resources in Anchor — usually the action step of a workflow, after the agent has gathered context. Every call changes real data in your Anchor environment.
The memory_manager tool creates and stores new memories/decisions in persistent state, which is a reversible data modification operation (Write). While it also supports Read operations (search, list), the presence of the add/create capability for memories elevates it beyond Read-only.
From the tool's definition Tool description states 'add a learning/decision' and manages memory state, which involves creating and modifying persistent data. Alongside 'search memories' and 'list recent memories' (Read operations), the add capability places this in Write category.
Attacks that exploit this kind of access
Manage memory: add a learning/decision, search memories, or list recent memories. It is categorised as a Write tool in the Anchor MCP Server, which means it can create or modify data. Consider rate limits to prevent runaway writes.
Register the Anchor MCP server in PolicyLayer and add a rule for memory_manager: 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 Anchor. Nothing to install.
memory_manager 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 memory_manager 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 memory_manager. 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.
memory_manager is provided by the Anchor MCP server (thewillmoss/anchor-mcp). 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 →