AI agents use kv_set to create or update resources in AgentOS — usually the action step of a workflow, after the agent has gathered context. Every call changes real data in your AgentOS environment.
The tool stores data in persistent key-value memory, which is a create/modify operation. It is reversible (data can be deleted via kv_delete or overwritten), so it falls under Write rather than Destructive. Severity is medium because misuse could corrupt agent state or memory, affecting workflow integrity, but the impact is limited to application state rather than system-wide or financial consequences.
From the tool's definition Tool name 'kv_set' and description 'Store a key-value pair' indicate data creation/modification in persistent memory. This is a write operation that creates or modifies data reversibly.
Attacks that exploit this kind of access
Store a key-value pair in persistent local memory with optional TTL and namespace. It is categorised as a Write tool in the AgentOS MCP Server, which means it can create or modify data. Consider rate limits to prevent runaway writes.
Register the AgentOS MCP server in PolicyLayer and add a rule for kv_set: 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 AgentOS. Nothing to install.
kv_set 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 kv_set 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 kv_set. 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.
kv_set is provided by the AgentOS MCP server (netflypsb/agentos). 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 →