DiCert MCP
DSN's MCP server that sits as an AI-DPI middleware layer over a DPI platform, exposing its data as secure, callable tools so AI workflows and voice agents can interact with DPI safely.
DiCert MCP is a Model Context Protocol (MCP) server that sits on top of a Digital Public Infrastructure (DPI) platform and exposes its data as secure, callable tools for AI workflows. It is the middleware pattern that lets AI clients – voice agents, chatbots, autonomous agents – interact with a DPI and its data through a governed interface rather than touching the raw database. The reference implementation runs over DiCert, Nigeria's credentialing platform for primary and secondary schools.
The pattern: AI-DPI middleware
DPI platforms hold sensitive national data behind their own schemas and access rules. Letting an AI model query them directly is neither safe nor practical. An MCP server solves this by acting as a standardised, governed layer between AI clients and the DPI:
- It connects to the DPI data store, normalises the data into clean shapes an LLM can use, and serves it over a standard MCP interface (HTTP / SSE).
- AI clients discover the available tools and call them – they never see raw credentials or issue arbitrary queries.
- Access can be gated with an API key, and the exposed surface is limited to purpose-built read tools rather than open database access.
This is exactly the interaction model NAISH wants for AI-on-DPI: capable, but bounded and auditable.
What DiCert MCP exposes
Over the DiCert credentialing data, the server offers tools such as:
| Tool | Purpose |
|---|---|
lookup_student / find_students | Look up or search students by ID, name, school, class, or year |
fetch_term_report | Full term report – subjects, attendance, psychomotor, behavioural, teacher comments |
fetch_student_history | Complete academic history across terms |
fetch_class_performance | Class-wide breakdown with per-subject averages, for teachers |
fetch_regional_result / search_regional_results | Regional / external examination certification records |
fetch_regional_school_summary | School-level pass and verification statistics |
get_ai_recommendation | AI-generated personalised feedback for a student, parent, or teacher |
Different AI personas call the same tools for different ends – a parent checking a child's results, a teacher reviewing class performance and spotting students needing intervention, an administrator pulling regional verification statistics.
Relevance to NAISH
DiCert MCP is the secure connective tissue between AI and DPI that the Hub's mission depends on. It shows how AI workflows can act on national data platforms without unsafe direct access – aligning with the Hub's AI Governance and Securing AI-Powered Solutions principles (bounded tool surfaces, API-key gating, normalised read views over raw data). It continues the AI-DPI lineage of the AI-DPI Sandbox: where the sandbox gave startups access to DPI building blocks, DiCert MCP gives AI agents a safe way to use them.
Built by Data Science Nigeria (DSN).
EdoDiDA Chatbot
DSN's Gates-funded natural-language chatbot built on DHIS2 for the Edo State Digital Governance and Data Management Agency, turning government data into plain-language insight for decision-makers.
AfricanVoices
DSN's AfricanVoices – a Gates-funded platform democratizing African speech data, with 1,900 hours of curated audio across Hausa, Igbo, Nigerian Pidgin, and Yoruba for voice-AI research.