Use the corpus
Same pattern as corpus.lantern.io: a hosted MCP server for zero-install queries, plus a local stdio server for offline/private use.
Hosted (recommended)
claude mcp add --transport http -s user geedge \ https://geedge.lantern.io/mcp
Self-hosted
go install github.com/getlantern/geedge/cmd/geedge-mcp@latest git clone https://github.com/getlantern/geedge ~/code/geedge claude mcp add -s user geedge \ $(go env GOPATH)/bin/geedge-mcp -- --corpus $HOME/code/geedge
Tools exposed
search_findings— full-text + tag-filter search over findings.get_finding— fetch a single finding by id.get_artifact— fetch a single artifact provenance record by id.list_taxonomy— the controlled vocabulary, so a caller knows what IDs to filter on.find_related— findings sharing censor/technique/product tags with a given one.synthesize— answer a question by retrieving every relevant finding with citations, for an LLM to compose from.
Visibility
The public endpoint serves only visibility: public findings — that's a hard data-path filter applied at build time, not a runtime check. Every record starts internal; promotion is a deliberate human step. See About for the review policy.