The Index · AI & Agents

DeepSeek vs Kage

These two are level on the tashan score (within 1 points), so the score does not decide it. The evidence rows below are where they differ. Both are measured the same way, on the same day, from public evidence only — so this is a comparison of two measurements rather than two marketing pages. How we measure ›

DeepSeekKagewhat it means
tashan score7069upkeep + freshness, gated by adoption
Adoption evidence3k/wk1k/wkthe raw public signal the score came from
Last releaseactiveactivenpm publish or repo push
Upkeep8173cadence, maintainers, status
Expertise gradesolidsolidan LLM read of its own documentation against a fixed rubric
Security auditno known advisories, no build provenanceno known advisories, no build provenanceexistence of findings is always free; the detail a licence buys is on each dossier
LicenceMIT
Maintainersmore than oneone primarybus-factor risk

Install either

DeepSeek

claude mcp add deepseek-mcp-server -- npx -y deepseek-mcp-server

Kage

claude mcp add kage-core-kage-graph-mcp -- npx -y @kage-core/kage-graph-mcp

Neither score is a security verdict — "well maintained" and "nothing known is wrong" are different claims, which is why the audit is a separate row and never folded into the number. The full audit for each, including what a licence adds, is on its own page: DeepSeek · Kage.

All ranked AI & Agents ›