PDDikti API & MCP
The Challenge
PDDikti is the national higher-education dataset, and it is the source of truth for institutions, study programmes, lecturers, and students. Consuming it programmatically is another matter: the upstream sits behind Cloudflare, the response shapes are undocumented, and every team that needs the data ends up writing the same brittle scraper. We wanted a version that could be dropped into an application — or handed to an AI agent — without the consumer needing to know any of that.
The Solution
We built a read-only proxy on Bun and Hono with every endpoint described by a Zod schema, which generates the OpenAPI document and the validation from the same definition — so the docs cannot drift from the behaviour. Cloudflare clearance is handled transparently by an integrated FlareSolverr solver, so consumers never see a challenge page. The same surface is exposed as a Model Context Protocol server, so an AI assistant can query institutions and programmes as tools. A React explorer SPA ships alongside for people who just want to look something up. It is public and self-hostable.
Zod-validated schema with generated docs
Also exposed as a Model Context Protocol server
Public repository, free to self-host
Designed & engineered by Manggala Code
