Compare page omits the tools people actually choose between: Vestige, CASS, ctx, PAXM

The compare page routes people to Mem0, Letta, memU, Memori, claude-mem, agentmemory, MemPalace, cognee/Graphiti and Hindsight. Most of those are a different category — SDK memory for applications you are building. The tools a person actually chooses between when they want their existing coding-agent history searched are missing: - **Vestige** — closest by philosophy: memory lifecycle, contradiction detection, supersede, decay. Local MCP server, AGPL, ships a ~130 MB embedding model. Publishes its own benchmark on causally-linked recall, which is exactly where a lexical index is weakest. Honest routing has to say so. - **CASS** — indexes existing session stores across a wide set of harnesses with a strong TUI, but does not turn them into automatic recall inside the agent. - **ctx** — Rust, local SQLite index over agent history, and it describes itself in almost our words: preserving and searching the reasoning history from prior sessions. - **PAXM** — provider-neutral adapter with passive hooks and resumable historical import. Each of these reads history that predates its own install, so the compare page cannot keep implying that is ours alone. What is ours is the combination: 17 harnesses, automatic recall, and no model or runtime. The page is titled an honest comparison, and that only stays true if it names the tools a reader is genuinely torn between — including the cases where the answer is "use theirs". The MemPalace entry already does this well; the same treatment for Vestige is the missing one. Facts should be dated and re-checked, since #1176 was exactly this going stale.