## 🧠 MemKraft — Memory API first MemKraft is installed locally. **Before editing `memory/*` files by hand, try the Python API first.** ### 6 core calls ```python from memkraft import MemKraft mk = MemKraft() # respects MEMKRAFT_DIR env, falls back to ./memory mk.track("Hong Gildong", entity_type="person", source="DM") mk.update("Hong Gildong", "joined Hashed as CTO", source="press") mk.search("Hashed CEO") mk.tier_set("hong-gildong", tier="core") # core / recall / archival mk.fact_add("Simon", "role", "CEO", valid_from="2020-03-01") mk.log_event("vibekai deploy done", tags="deploy", importance="high") ``` ### Gotchas - Tier values are `core` / `recall` / `archival` **only** (`critical` ❌). - `decay_rate` is in (0, 1) exclusive — not `weight`. - `promote()` (markdown tag) ≠ `tier_set()` (frontmatter). Prefer `tier_set`. - Past memory lookup → `mk.search(...)` before `grep`. - After editing `[[wiki-links]]`, call `mk.link_scan()`. ### Regenerate this block ```bash memkraft agents-hint claude-code > AGENTS.md # or paste into an existing file ```