keyword_search
AI agents call keyword_search to retrieve information from Obsidian MCP (pgvector + Ollama, self-hosted) without modifying anything — typically the context-gathering step in research, monitoring, and reporting workflows, before the agent takes action elsewhere.
keyword_search retrieves data via text search with no side effects. It matches the Read category pattern. Severity is medium rather than low because: (1) full-text search over a personal knowledge vault could expose sensitive notes if an adversarial prompt tricks the tool into returning private content, and (2) the self-hosted PostgreSQL backend is a potential data exfiltration vector if search results are not…
From the tool's definition Tool name 'keyword_search' indicates a search/retrieval operation. Server description mentions 'semantic + full-text search over PostgreSQL/pgvector' as a core function, and 'keyword_search' logically fits this capability.
Attacks that exploit this kind of access
keyword_search. It is categorised as a Read tool in the Obsidian MCP (pgvector + Ollama, self-hosted) MCP Server, which means it retrieves data without modifying state.
Register the Obsidian MCP (pgvector + Ollama, self-hosted) MCP server in PolicyLayer and add a rule for keyword_search: 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 Obsidian MCP (pgvector + Ollama, self-hosted). Nothing to install.
keyword_search 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 keyword_search 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 keyword_search. 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.
keyword_search is provided by the Obsidian MCP (pgvector + Ollama, self-hosted) MCP server (maxkuminov/obsidian-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 →