AMCS is a Go MCP server for capturing project thoughts, semantic retrieval, summaries, and linked memory workflows with Postgres + pgvector.
It stores durable memory for assistants, supports project scoping, and exposes tools over MCP for capture, search, context recall, and structured operations.
Server status
{#if loading}Loading…
{:else if error}Unavailable
{error}
{:else if data}{data.connected_count}
connected in {data.connected_window}
{#if data.entries.length > 0}
Clients:
{#each data.entries.slice(0, 3) as client, idx}
{client.key_id}{idx <
Math.min(data.entries.length, 3) - 1
? ", "
: ""}
{/each}
{data.entries.length > 3 ? " …" : ""}
No status snapshot yet.
{/if}Memory stack
Postgres + pgvector
Semantic search with full-text fallback when vectors are missing.
AMCS separates reusable behavior, safety constraints, and curated knowledge so assistants can be guided consistently across sessions.