flowchart LR
subgraph SRC[" SOURCES — 17 OPEN-SOURCE TOOLKITS "]
direction TB
BC["Bellingcat Toolkit"]
AO["Awesome OSINT"]
OFW["OSINT Framework"]
MORE["+ 14 more"]
end
CRW["Weekly crawler
new · changed · dead"]
ENR["Enrichment
categories · capabilities · status"]
IDX[("9,000+ tools
fresh embeddings weekly")]
RAG["RAG + semantic query
grounded answers · cached Q&A"]
AGT(["Your agent
MCP · Claude Code · Codex · Goose"])
WEB(["Your browser
WebGPU · fine-tuned model · local"])
COM["Community
notes · votes · missing tools · reports"]
BC --> CRW
AO --> CRW
OFW --> CRW
MORE --> CRW
CRW --> ENR --> IDX --> RAG
RAG ==> AGT
RAG ==> WEB
AGT -. "votes · submissions" .-> COM
WEB -. "votes · submissions" .-> COM
COM -. "Wilson-ranked, back into the index" .-> IDX
classDef source fill:#D4A85314,stroke:#D4A853,stroke-width:1.5px
classDef pipe fill:#F5F2ED0a,stroke:#8b939c,stroke-width:1.5px
classDef index fill:#D4A85330,stroke:#D4A853,stroke-width:2px
classDef users fill:#5B8A8A26,stroke:#5B8A8A,stroke-width:1.5px
classDef community fill:#4ade8016,stroke:#4ade80,stroke-width:1.5px
class BC,AO,OFW,MORE source
class CRW,ENR,RAG pipe
class IDX index
class AGT,WEB users
class COM community
Navigator works inside Spotlight and inside members' existing workflows. Meeting researchers in the agents they already use did more for adoption than any interface.
Upstream contributions keep researchers aware of changes in tooling — and keep maintainers updated. The useful layer comes from people pushing on it in real investigations.
RAG answers are only as good as the metadata behind them. And local inference proved to be a trust feature, not a performance one.