Find similar vectors using cosine, L2, or inner product distance (requires pgvector)
AI agents call vector_similarity to retrieve information from Postgres Mcp Legacy without modifying anything — typically the context-gathering step in research, monitoring, and reporting workflows, before the agent takes action elsewhere.
The tool performs a similarity search against vectors, which is a read-only query operation. It retrieves data based on distance calculations but does not create, modify, delete, or execute code. The pgvector dependency is a storage/indexing detail. Severity is low because misuse (e.g., querying sensitive embeddings) poses minimal risk compared to data corruption or code execution.
From the tool's definition Tool name and description indicate 'Find similar vectors' — a retrieval/query operation that measures distance between vectors using standard metrics (cosine, L2, inner product). No modification, deletion, execution, or financial action is described.
Documented attack patterns abuse exactly the kind of access vector_similarity gives an agent:
PolicyLayer is an MCP gateway — it sits between your AI agents and Postgres Mcp Legacy, and nothing reaches the server without passing your rules. This is the rule we recommend for vector_similarity:
{
"version": "1",
"default": "deny",
"tools": {
"vector_similarity": {}
}
}vector_similarity is read-only, so it stays allowed — but everything else on the server is denied unless you say otherwise.
Free to start. No card required.
Find similar vectors using cosine, L2, or inner product distance (requires pgvector). It is categorised as a Read tool in the Postgres Mcp Legacy MCP Server, which means it retrieves data without modifying state.
Register the Postgres Mcp Legacy MCP server in PolicyLayer and add a rule for vector_similarity: 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 Postgres Mcp Legacy. Nothing to install.
vector_similarity is a Read tool with low risk. Read-only tools are generally safe to allow by default.
Yes. Add a rate_limit block to the vector_similarity 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 vector_similarity. 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.
vector_similarity is provided by the Postgres Mcp Legacy MCP server (neverinfamous/postgres-mcp-legacy). PolicyLayer sits as a proxy in front of this server to enforce policies before tool calls reach the server.
Start from Postgres Mcp Legacy, add the rest of your stack, and see everything your agents can call. Then put policy on all of it.
Free to start. No card required.
60 Postgres Mcp Legacy tools catalogued and risk-classified — across an index of 43,000+ MCP servers.