What is Driver?
Driver pre-computes deep, structured documentation for your codebases and makes it available to AI coding agents and humans.
The problem
AI coding agents fail on complex tasks because they lack codebase context. This breaks down in three specific ways:
They miss critical information. Agents search for relevant code at runtime using grep, file search, and embeddings. These approaches work for small tasks but fall off a cliff as codebases grow. Critical edge cases, transitive dependencies, and architectural constraints get missed entirely.
They can't separate signal from noise. What counts as relevant context changes with every task. Information useful for a refactor can cause hallucinations during a bug fix. Naive retrieval methods flood the context window with structurally similar but semantically wrong code.
They get the wrong level of detail. Complex tasks need different context at different stages. Early planning requires high-level architecture. Implementation needs specific function signatures and conventions. Current tools fetch context at a single abstraction level and hope it fits.
What Driver generates
For every connected codebase, Driver produces:
- Architecture overviews: system structure, component relationships, key design decisions
- Code maps: every file and directory annotated with descriptions
- File-level documentation: functions, classes, interfaces, with symbol-level detail
- Changelogs: derived from git history, organized by month
- Onboarding guides: optimized for AI agents to get oriented fast
Learn more about the Transpiler and its approach in How Driver Works. Driver keeps these artifacts current as your codebase evolves.
How you access it
Driver serves context through MCP (Model Context Protocol) directly in your IDE. Supported editors:
Your AI agent queries Driver's MCP server using 7 tools to retrieve architecture, file documentation, and history on demand. No copy-pasting, no stale wikis.