An AI agent that scans, tags, and triages every government PDF — reading the catalog through the DataHub MCP Server and writing every verdict back. It runs every day, so people don't have to.
git clone https://github.com/imeunk89/accesscatalog-agent && cd accesscatalog-agent uv venv --python 3.11 && source .venv/bin/activate uv pip install -e . mcp-server-datahub datahub docker quickstart # local DataHub (Docker) cp .env.example .env # add your OPENAI_API_KEY ./scripts/demo.sh # ingest → agent → reports
The 22-PDF sample corpus ships in the repo — genuinely accessible and inaccessible files, including image-only scans and lineage-connected remediation pairs.