valv

Product Introduction

1. Definition: Valv is a secure data access and query platform designed for AI agents. Technically, it is a Model Context Protocol (MCP) server that acts as a middleware layer between AI assistants (like Claude Code) and an organization's live databases, including Postgres, MySQL, ClickHouse, and PostHog.
2. Core Value Proposition: Valv exists to enable teams to safely grant AI agents direct, natural language query access to live production data. Its primary value is enforcing granular, row-level security guardrails, ensuring agents only retrieve data explicitly permitted for their assigned role, thereby mitigating data exposure risks.

Main Features

1. Secure Database Connector: Valv establishes authenticated connections to multiple database types (Postgres, MySQL, ClickHouse, PostHog) through a single interface. It uses OAuth for secure MCP server registration and manages connection pooling and credential security, abstracting the complexity of direct database links from the AI agent.
2. Natural Language to SQL Translation with Guardrails: The core functionality allows AI agents to query connected databases using plain English. The system interprets the agent's intent, but crucially, all queries are executed within a strictly defined security context. This means every generated SQL query is automatically filtered and scoped according to pre-configured row-level access policies before execution.
3. Role-Based Row-Level Security (RLS): This is the defining technical feature. Administrators define roles (e.g., "support_agent_west_coast," "marketing_analyst_q2") and attach granular data access rules to each. These rules, enforced at the database query level, ensure that any agent using that role can only see rows of data they are authorized to view, such as customer tickets only from a specific region or sales data only for a certain timeframe.

Problems Solved

1. Pain Point: The conflict between leveraging AI agent productivity and maintaining stringent data security. Manually fetching data for agents is slow, while giving agents direct database access poses severe compliance and privacy risks from over-permissioned queries.
2. Target Audience: Data Engineers & DevOps Engineers who need to provide data access securely; Product Teams & Developers building AI-augmented features; Analytics & Marketing Teams who want self-service data querying without SQL expertise; Security & Compliance Officers responsible for data governance.
3. Use Cases: A customer support AI agent can query live ticket data but is restricted to only the tickets for the customer it is assisting. A marketing analyst can ask an AI to "compare Q3 conversion rates by campaign" and get an answer without seeing underlying PII. A developer's AI coding assistant can query schema and anonymized production data to debug issues without exposing sensitive user information.

Unique Advantages

1. Differentiation: Unlike generic database clients or BI tools, Valv is built specifically for the AI agent workflow with security as the foundation. Compared to manually crafting API endpoints for AI access, Valv provides dynamic, natural language querying. It differs from simply using a database's native RLS by centralizing policy management across multiple data sources and integrating seamlessly into the MCP ecosystem used by modern AI assistants.
2. Key Innovation: The integration of the Model Context Protocol (MCP) as the transport layer. This allows Valv to function as a standardized, pluggable "tool" for AI agents like Claude, turning it into a universally compatible data gateway. The innovation lies in combining this agent-native protocol with enterprise-grade, query-level security enforcement.

Frequently Asked Questions (FAQ)

1. How does Valv ensure my production data is safe from AI agents? Valv enforces safety by intercepting every natural language query, converting it to SQL, and dynamically applying pre-configured row-level security (RLS) filters before the query is executed on your database. The AI agent only receives the filtered, authorized result set.
2. What databases are compatible with the Valv platform? Valv currently provides native integration and secure connectivity for PostgreSQL, MySQL, ClickHouse, and PostHog databases, allowing you to manage access across these systems from a single control plane.
3. What is the Model Context Protocol (MCP) and why does Valv use it? The Model Context Protocol (MCP) is an open standard that allows AI assistants to securely connect to external data sources and tools. Valv uses MCP to function as a standardized server, enabling seamless and authenticated integration with AI agent clients like Claude Code without requiring custom, fragile integrations.
4. Can I use Valv with AI models other than Anthropic's Claude? Yes, while the provided onboarding references Claude Code, Valv is built on the open MCP standard. Any AI assistant or development environment that supports MCP clients (including potential future integrations with other AI platforms) can theoretically connect to a Valv server to enable secure database querying.