
by valv
valv gives AI agents structured tools to read a database, and optionally write to it, with access limited by policies you write in code. The model never sends native database commands.
AgentContact for pricing
Step 1: Install an adapter for your database with npm.
Step 2: Call createValv with your database client and a deny-all default policy.
Step 3: Write policies for each resource using your request context.
Step 4: Generate tools for your agent framework and add the valv instructions to the system prompt.
Step 5: Run the agent and let it discover and query allowed resources.
Step 6: Alternatively, point the MCP package at a database so a coding agent can query it.
It lists adapters for ClickHouse, Postgres, MySQL and SQLite through Prisma, and MongoDB.
Only if you turn it on. Write tools are off by default and write permissions default to denied.
No. Denied and unknown columns fail with the same message, so the model cannot probe for them.
Yes. The same tools can be exposed over MCP, or a coding agent can use the MCP package against a database.
MemoriMemori is a memory layer for AI agents that stores and recalls facts from conversations and agent actions. It ships Python and TypeScript SDKs, a cloud service and an MCP endpoint.
openITCOCKPIT MCPAn MCP server that lets an LLM client query an openITCOCKPIT monitoring instance and, if you enable write tools, act on it. It exposes 42 tools.