Code Fundi

Code Fundi is an AI search tool. A coding AI that indexes repositories into a token-optimized codebase map, adds blast-radius analysis of code changes, multi-repo semantic search and MCP or API access for coding agents, with editor extensions for VS Code, Cursor and Windsurf.

Code Fundi is a coding AI built around a searchable map of your codebase rather than a plain text assistant. It ingests a repository (via a pasted URL or a GitHub connection), maps the dependency graph, strips boilerplate and produces a token-optimized, structure-aware map of the project. That map becomes the context layer for chat, semantic search and external agents, so an assistant editing one file can see which other files, functions and services depend on it before the change ships. The headline feature is Blast-Radius Guard, which traces structure and flow instead of matching text, and is positioned against the silent regressions that occur when an agent edits a file without knowing what else breaks. Universal Search runs sub-second queries across multiple public repositories at once, and the same index backs single and multi-codebase chat. Outputs include markdown codebase data that can feed retrieval pipelines, documentation and prompt engineering, plus a training data factory that exports file-by-file logic into datasets for fine-tuning custom models. An agentic data layer exposes the map through MCP and an API so parallel agent teams can handle QA, documentation and refactoring work. Practically, the tool is consumed in two places: a web app for indexing, search and chat, and editor extensions for VS Code, Cursor and Windsurf, so much of the work happens where code is already written. It suits engineering teams running agentic workflows at scale, platform and DevEx groups who need grounded context for tools like Claude Code or Devin, and teams that want fine-tuning data from their own repositories. Compared with general chat models it is narrower and repository-centric, and compared with text-similarity retrieval tools it emphasizes dependency structure and change impact rather than chunk matching. Pricing is credit-based and tiers scale by private repository count, credit allowance and log retention.