OpenDB Ai

OpenDB is a local-first infrastructure suite designed for AI agents operating directly on a user's machine. It addresses the need for private, on-device agent capabilities by providing a context and runtime environment that keeps files, search, memory, and execution state local by default. This approach emphasizes privacy and provenance, ensuring that agent data such as files, process state, permissions, diffs, and review history remain on the developer's machine unless explicitly moved elsewhere.

The suite consists of two main products: openDB and openRuntime. openDB serves as an AI-native local database that agents can use for managing private files, performing search operations, and maintaining durable memory. It provides a context layer that agents can query before initiating tasks, which helps reduce unnecessary token usage and limits access to unrelated files. openRuntime functions as a control plane for coding agents, supporting the execution of tasks with features such as visible state tracking, approval gates, worktree isolation, and diff-first review. It enables the running of coding tasks—including those for Codex, Claude Code, and shell operations—with mechanisms for supervision and review.

The infrastructure employs simple and familiar primitives, including SQLite, FTS5, MCP, git worktrees, and small APIs, to offer foundational capabilities before introducing more complex infrastructure. The design philosophy prioritizes evidence and transparency, making search results, memories, events, and diffs easily traceable for users. The system is zero-configuration and does not require a remote runtime, reinforcing its local-first orientation.

OpenDB is positioned as a product family within the broader class of local agent infrastructure tools, specifically catering to developers who require private, supervised, and reviewable execution environments for their AI agents.