memrust 0.6.1

Agent-native memory engine: hybrid retrieval (HNSW + BM25 + entity graph + recency) behind remember/recall/forget, with HTTP and MCP interfaces
Documentation
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
<!doctype html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>memrust — memory infrastructure for AI agents</title>
<meta name="description" content="An agent-native memory engine in Rust. Hybrid retrieval — vectors, keywords, an entity graph and recency — behind remember/recall/forget, with every result explaining why it surfaced.">
<script>
  try { var t = localStorage.getItem("memrust-site-theme"); if (t) document.documentElement.setAttribute("data-theme", t); } catch (e) {}
</script>
<style>
  /* ------------------------------------------------------------------
     memrust.dev — the four retrieval signals are the design system.
     vector / lexical / graph / recency carry the same hues here that
     they carry in the product's own dashboard.
     ------------------------------------------------------------------ */
  :root {
    color-scheme: dark;
    --ink:        #f4f1ec;
    --ink-2:      #b3ada3;
    --ink-3:      #7d776e;
    --ground:     #0d0b0a;   /* near-black, warmed toward the brand hue */
    --ground-2:   #151211;
    --panel:      #191614;
    --rail:       #2b2622;
    --rail-soft:  #201c19;
    --brand:      #f97316;
    --brand-ink:  #fb9a4e;
    --sig-vector: #3987e5;
    --sig-lexical:#d95926;
    --sig-graph:  #199e70;
    --sig-recency:#c98500;
    --shell: min(1180px, 100% - 3rem);
    --step--1: clamp(.75rem, .72rem + .12vw, .82rem);
    --step-0:  clamp(1rem, .96rem + .2vw, 1.09rem);
    --step-1:  clamp(1.2rem, 1.1rem + .5vw, 1.5rem);
    --step-2:  clamp(1.6rem, 1.3rem + 1.4vw, 2.4rem);
    --step-3:  clamp(2.1rem, 1.55rem + 2.2vw, 3.55rem);
    --mono: ui-monospace, SFMono-Regular, "SF Mono", Menlo, Consolas, monospace;
    --sans: ui-sans-serif, -apple-system, BlinkMacSystemFont, "Segoe UI", Roboto, "Helvetica Neue", Arial, sans-serif;
  }
  @media (prefers-color-scheme: light) {
    :root:where(:not([data-theme="dark"])) {
      color-scheme: light;
      --ink: #17130f; --ink-2: #554e46; --ink-3: #857d73;
      --ground: #f7f4ef; --ground-2: #efeae2; --panel: #ffffff;
      --rail: #ded7cc; --rail-soft: #e8e2d8;
      --brand: #ea580c; --brand-ink: #c2410c;
      --sig-vector: #2a78d6; --sig-lexical: #c2410c; --sig-graph: #17845f; --sig-recency: #9a6a00;
    }
  }
  :root[data-theme="light"] {
    color-scheme: light;
    --ink: #17130f; --ink-2: #554e46; --ink-3: #857d73;
    --ground: #f7f4ef; --ground-2: #efeae2; --panel: #ffffff;
    --rail: #ded7cc; --rail-soft: #e8e2d8;
    --brand: #ea580c; --brand-ink: #c2410c;
    --sig-vector: #2a78d6; --sig-lexical: #c2410c; --sig-graph: #17845f; --sig-recency: #9a6a00;
  }

  * { box-sizing: border-box; }
  html { scroll-behavior: smooth; }
  @media (prefers-reduced-motion: reduce) { html { scroll-behavior: auto; } *, *::before, *::after { animation: none !important; transition: none !important; } }
  body {
    margin: 0; background: var(--ground); color: var(--ink);
    font-family: var(--sans); font-size: var(--step-0); line-height: 1.6;
    -webkit-font-smoothing: antialiased;
  }
  a { color: inherit; }
  ::selection { background: var(--brand); color: #16100a; }
  :focus-visible { outline: 2px solid var(--brand); outline-offset: 3px; border-radius: 3px; }

  /* The framed console: rails down both sides, ticks at each section seam. */
  .frame { width: var(--shell); margin-inline: auto; border-inline: 1px solid var(--rail-soft); position: relative; }
  .seam { border-top: 1px solid var(--rail); position: relative; }
  .seam::before, .seam::after {
    content: ""; position: absolute; top: -3px; width: 5px; height: 5px; background: var(--brand);
  }
  .seam::before { left: -3px; } .seam::after { right: -3px; }
  .pad { padding: clamp(3rem, 7vw, 6.5rem) clamp(1.1rem, 3.4vw, 3rem); }
  section[id] { scroll-margin-top: 4.5rem; }

  .eyebrow {
    font-family: var(--mono); font-size: var(--step--1); letter-spacing: .14em;
    text-transform: uppercase; color: var(--brand-ink); margin: 0 0 1.4rem;
  }
  h1, h2, h3 { font-weight: 800; letter-spacing: -.025em; text-wrap: balance; margin: 0; line-height: 1.03; }
  h1 { font-size: var(--step-3); }
  h2 { font-size: var(--step-2); line-height: 1.08; }
  h3 { font-size: var(--step-1); letter-spacing: -.015em; }
  .accent { color: var(--brand); font-style: italic; font-weight: 800; }
  p { margin: 0; }
  .lede { color: var(--ink-2); font-size: var(--step-1); line-height: 1.5; max-width: 40ch; font-weight: 400; letter-spacing: -.01em; }
  .prose { color: var(--ink-2); max-width: 62ch; }
  .mono { font-family: var(--mono); }
  .tnum { font-variant-numeric: tabular-nums; }

  /* header ---------------------------------------------------------- */
  header { position: sticky; top: 0; z-index: 20; background: color-mix(in srgb, var(--ground) 88%, transparent); backdrop-filter: blur(10px); border-bottom: 1px solid var(--rail-soft); }
  .bar { width: var(--shell); margin-inline: auto; display: flex; align-items: center; gap: .85rem; padding: .7rem clamp(1.1rem, 3.4vw, 3rem); }
  .wordmark { display: flex; align-items: center; gap: .55rem; font-weight: 800; letter-spacing: -.02em; font-size: 1.05rem; text-decoration: none; }
  .chip {
    font-family: var(--mono); font-size: .68rem; letter-spacing: .08em; color: var(--ink-3);
    border: 1px solid var(--rail); border-radius: 4px; padding: .16rem .42rem;
  }
  nav { display: none; gap: 1.4rem; margin-inline: auto; font-size: .92rem; }
  nav a { color: var(--ink-2); text-decoration: none; }
  nav a:hover { color: var(--ink); }
  @media (min-width: 900px) { nav { display: flex; } }
  .btn {
    display: inline-flex; align-items: center; gap: .45rem; text-decoration: none;
    font-size: .9rem; font-weight: 600; padding: .5rem .9rem; border-radius: 7px;
    border: 1px solid var(--rail); color: var(--ink); background: var(--ground-2);
    transition: border-color .15s, background .15s;
  }
  .btn:hover { border-color: var(--ink-3); }
  .btn-brand { background: var(--brand); border-color: var(--brand); color: #180f07; }
  .btn-brand:hover { background: var(--brand-ink); border-color: var(--brand-ink); }
  .icon-btn { padding: .45rem .55rem; line-height: 0; }
  .icon-btn svg { width: 16px; height: 16px; stroke: currentColor; fill: none; stroke-width: 1.9; stroke-linecap: round; stroke-linejoin: round; }
  .i-sun { display: inline; } .i-moon { display: none; }
  :root[data-theme="light"] .i-sun { display: none; } :root[data-theme="light"] .i-moon { display: inline; }
  @media (prefers-color-scheme: light) {
    :root:where(:not([data-theme="dark"])) .i-sun { display: none; }
    :root:where(:not([data-theme="dark"])) .i-moon { display: inline; }
  }

  /* hero ------------------------------------------------------------ */
  .hero { display: grid; gap: clamp(2.2rem, 4vw, 3.4rem); }
  @media (min-width: 960px) { .hero { grid-template-columns: 1.12fr .88fr; align-items: center; } }
  .cta-row { display: flex; flex-wrap: wrap; gap: .7rem; margin-top: 2rem; }
  .facts { display: flex; flex-wrap: wrap; gap: .5rem 1.1rem; margin-top: 1.8rem; font-family: var(--mono); font-size: var(--step--1); color: var(--ink-3); }
  .facts span::after { content: "·"; margin-left: 1.1rem; color: var(--rail); }
  .facts span:last-child::after { content: ""; }

  /* the live recall card — the product's actual output ---------------- */
  .card { background: var(--panel); border: 1px solid var(--rail); border-radius: 12px; overflow: hidden; }
  .card-top {
    display: flex; align-items: center; justify-content: space-between; gap: 1rem;
    padding: .6rem .9rem; border-bottom: 1px solid var(--rail);
    font-family: var(--mono); font-size: var(--step--1); color: var(--ink-3);
  }
  .hit { padding: .85rem .9rem; border-bottom: 1px solid var(--rail-soft); }
  .hit:last-child { border-bottom: 0; }
  .hit-head { display: flex; align-items: baseline; gap: .55rem; flex-wrap: wrap; font-family: var(--mono); font-size: var(--step--1); }
  .score { color: var(--brand-ink); font-weight: 700; }
  .kind { color: var(--ink-3); border: 1px solid var(--rail); border-radius: 999px; padding: 0 .45rem; font-size: .68rem; }
  .hit-text { margin: .45rem 0 .6rem; font-size: .92rem; line-height: 1.45; }
  .sigs { display: grid; grid-template-columns: repeat(2, 1fr); gap: .55rem .9rem; }
  @media (min-width: 560px) { .sigs { grid-template-columns: repeat(4, 1fr); } }
  .sig-name { font-family: var(--mono); font-size: .64rem; color: var(--ink-3); letter-spacing: .04em; display: flex; justify-content: space-between; gap: .3rem; }
  .track { height: 5px; border-radius: 3px; background: color-mix(in srgb, var(--ink) 10%, transparent); margin-top: .22rem; overflow: hidden; }
  .fill { height: 100%; border-radius: 3px; transform-origin: left; animation: grow .9s cubic-bezier(.2,.7,.3,1) both; }
  @keyframes grow { from { transform: scaleX(0); } to { transform: scaleX(1); } }
  .v { background: var(--sig-vector); } .l { background: var(--sig-lexical); }
  .g { background: var(--sig-graph); } .r { background: var(--sig-recency); }
  .card-note { padding: .6rem .9rem; border-top: 1px solid var(--rail); font-family: var(--mono); font-size: var(--step--1); color: var(--ink-3); }
  .card-note b { color: var(--sig-graph); font-weight: 600; }

  /* the stack ------------------------------------------------------- */
  .layer-wrap { display: grid; gap: 2.6rem; }
  @media (min-width: 940px) { .layer-wrap { grid-template-columns: 1.05fr .95fr; align-items: center; gap: 4rem; } }
  .stack { display: flex; flex-direction: column; gap: .5rem; }
  .tier {
    position: relative; border: 1px solid var(--rail-soft); border-radius: 9px;
    background: var(--ground-2); padding: 1rem 1.1rem; display: flex; flex-direction: column; gap: .2rem;
  }
  .tier-l { font-weight: 700; font-size: .98rem; }
  .tier-v { font-family: var(--mono); font-size: var(--step--1); color: var(--ink-3); }
  .tier-me {
    border-color: var(--brand); background: color-mix(in srgb, var(--brand) 11%, var(--ground-2));
    padding-block: 1.35rem;
  }
  .tier-me .tier-l { color: var(--brand); font-size: 1.15rem; }
  .tier-badge {
    position: absolute; top: 50%; right: 1.1rem; transform: translateY(-50%);
    font-family: var(--mono); font-size: var(--step--1); font-weight: 600;
    background: var(--brand); color: #180f07; border-radius: 999px; padding: .2rem .6rem;
  }

  /* the 27% wall ----------------------------------------------------- */
  .wall { display: grid; gap: 2.4rem; }
  @media (min-width: 900px) { .wall { grid-template-columns: 1fr 1.15fr; align-items: start; gap: 3.5rem; } }
  .bars { display: flex; flex-direction: column; gap: 1rem; }
  .bar-row { display: grid; grid-template-columns: 1fr; gap: .3rem; }
  .bar-label { display: flex; justify-content: space-between; align-items: baseline; gap: 1rem; font-family: var(--mono); font-size: var(--step--1); color: var(--ink-2); }
  .bar-label b { font-size: 1.05rem; color: var(--ink); font-weight: 700; }
  .bar-track { height: 30px; background: var(--ground-2); border: 1px solid var(--rail-soft); border-radius: 5px; overflow: hidden; }
  .bar-fill { height: 100%; background: color-mix(in srgb, var(--sig-vector) 55%, var(--ground)); border-right: 2px solid var(--sig-vector); }
  .bar-fill.win { background: color-mix(in srgb, var(--brand) 40%, var(--ground)); border-right-color: var(--brand); }
  .bar-row.muted .bar-label { color: var(--ink-3); }

  /* stats ------------------------------------------------------------ */
  .stats { display: grid; grid-template-columns: repeat(auto-fit, minmax(200px, 1fr)); }
  .stat { padding: 1.6rem clamp(1.1rem, 3vw, 2rem); border-top: 1px solid var(--rail-soft); border-right: 1px solid var(--rail-soft); }
  .stat-v { font-size: clamp(2rem, 1.2rem + 2.6vw, 3.2rem); font-weight: 800; letter-spacing: -.045em; line-height: 1; }
  .stat-v small { font-size: .38em; font-weight: 600; letter-spacing: -.01em; color: var(--ink-3); margin-left: .25em; }
  .stat-l { font-family: var(--mono); font-size: var(--step--1); color: var(--ink-3); margin-top: .6rem; line-height: 1.45; }

  /* signal grid ------------------------------------------------------ */
  .grid-4 { display: grid; gap: 1px; background: var(--rail-soft); border: 1px solid var(--rail-soft); border-radius: 10px; overflow: hidden; }
  @media (min-width: 760px) { .grid-4 { grid-template-columns: repeat(2, 1fr); } }
  @media (min-width: 1060px) { .grid-4 { grid-template-columns: repeat(4, 1fr); } }
  .cell { background: var(--ground); padding: 1.5rem 1.3rem; display: flex; flex-direction: column; gap: .55rem; }
  .dot { width: 9px; height: 9px; border-radius: 50%; }
  .cell-h { display: flex; align-items: center; gap: .5rem; font-family: var(--mono); font-size: var(--step--1); letter-spacing: .1em; text-transform: uppercase; color: var(--ink-3); }
  .cell p { color: var(--ink-2); font-size: .93rem; }

  /* table ------------------------------------------------------------ */
  .scroller { overflow-x: auto; border: 1px solid var(--rail-soft); border-radius: 10px; }
  table { border-collapse: collapse; width: 100%; min-width: 640px; font-size: .9rem; }
  th, td { text-align: left; padding: .72rem .9rem; border-bottom: 1px solid var(--rail-soft); }
  thead th { font-family: var(--mono); font-size: var(--step--1); font-weight: 500; letter-spacing: .08em; text-transform: uppercase; color: var(--ink-3); background: var(--ground-2); }
  tbody tr:last-child td { border-bottom: 0; }
  td.num { font-family: var(--mono); font-variant-numeric: tabular-nums; }
  .me td { background: color-mix(in srgb, var(--brand) 9%, transparent); }
  .me td:first-child { font-weight: 700; }
  .yes { color: var(--sig-graph); font-weight: 700; }
  .partial { color: var(--sig-recency); }
  .no { color: var(--ink-3); }

  /* code ------------------------------------------------------------- */
  .tabs { display: flex; gap: .35rem; flex-wrap: wrap; margin-bottom: .9rem; }
  .tab {
    font-family: var(--mono); font-size: var(--step--1); padding: .35rem .75rem; border-radius: 999px; cursor: pointer;
    border: 1px solid var(--rail); background: transparent; color: var(--ink-2);
  }
  .tab[aria-selected="true"] { background: var(--brand); border-color: var(--brand); color: #180f07; font-weight: 600; }
  pre { margin: 0; padding: 1.1rem 1.2rem; overflow-x: auto; font-family: var(--mono); font-size: .84rem; line-height: 1.7; }
  pre .c { color: var(--ink-3); } pre .k { color: var(--sig-vector); } pre .s { color: var(--sig-graph); } pre .f { color: var(--brand-ink); }
  [hidden] { display: none !important; }

  /* faq --------------------------------------------------------------- */
  .faq { border: 1px solid var(--rail-soft); border-radius: 10px; overflow: hidden; }
  .faq details { border-bottom: 1px solid var(--rail-soft); background: var(--ground); }
  .faq details:last-child { border-bottom: 0; }
  .faq summary {
    cursor: pointer; list-style: none; padding: 1.05rem 1.25rem;
    display: flex; align-items: baseline; justify-content: space-between; gap: 1.5rem;
    font-weight: 650; font-size: 1.02rem; letter-spacing: -.012em;
  }
  .faq summary::-webkit-details-marker { display: none; }
  .faq summary::after {
    content: "+"; font-family: var(--mono); color: var(--brand);
    font-size: 1.15rem; line-height: 1; flex: none;
  }
  .faq details[open] summary::after { content: ""; }
  .faq summary:hover { color: var(--brand); }
  .faq .a { padding: 0 1.25rem 1.25rem; max-width: 72ch; display: flex; flex-direction: column; gap: .8rem; }
  .faq .a p { color: var(--ink-2); font-size: .95rem; }

  /* roadmap + contact ------------------------------------------------- */
  .road { display: grid; gap: 1px; background: var(--rail-soft); border: 1px solid var(--rail-soft); border-radius: 10px; overflow: hidden; margin-top: 2.4rem; }
  @media (min-width: 820px) { .road { grid-template-columns: repeat(3, 1fr); } }
  .road-item { background: var(--ground); padding: 1.4rem 1.3rem; display: flex; flex-direction: column; gap: .5rem; }
  .road-t { font-family: var(--mono); font-size: var(--step--1); letter-spacing: .1em; text-transform: uppercase; color: var(--brand-ink); }
  .road-item p { color: var(--ink-2); font-size: .93rem; }

  .contact-wrap { display: grid; gap: 2.6rem; }
  @media (min-width: 940px) { .contact-wrap { grid-template-columns: 1fr 1fr; gap: 4rem; align-items: center; } }
  .contact-cards { display: flex; flex-direction: column; gap: .6rem; }
  .contact-card {
    display: flex; flex-direction: column; gap: .15rem; text-decoration: none;
    border: 1px solid var(--rail-soft); border-radius: 10px; padding: 1.1rem 1.2rem;
    background: var(--ground-2); transition: border-color .15s, transform .15s;
  }
  .contact-card:hover { border-color: var(--brand); transform: translateX(3px); }
  .cc-k { font-family: var(--mono); font-size: var(--step--1); letter-spacing: .08em; text-transform: uppercase; color: var(--ink-3); }
  .cc-v { font-weight: 700; font-size: 1.02rem; }
  .contact-card p { color: var(--ink-2); font-size: .89rem; margin-top: .3rem; }

  /* footer ----------------------------------------------------------- */
  footer { color: var(--ink-3); font-family: var(--mono); font-size: var(--step--1); }
  .foot { display: flex; flex-wrap: wrap; gap: 1rem 2rem; align-items: center; justify-content: space-between; }
  .foot a { text-decoration: none; color: var(--ink-2); }
  .foot a:hover { color: var(--brand); }
  .foot-links { display: flex; flex-wrap: wrap; gap: 1.2rem; }
</style>
</head>
<body>

<header>
  <div class="bar">
    <a class="wordmark" href="#top">
      <svg width="26" height="26" viewBox="0 0 48 48" aria-hidden="true">
        <path d="M24 4 42 14v8L24 12 6 22v-8L24 4z" fill="#f97316"/>
        <rect x="12" y="22" width="24" height="7" rx="2" fill="currentColor" opacity=".8"/>
        <rect x="12" y="31" width="24" height="6" rx="2" fill="#f97316" opacity=".8"/>
        <rect x="12" y="39" width="24" height="5" rx="2" fill="currentColor" opacity=".35"/>
      </svg>
      memrust
    </a>
    <nav>
      <a href="#layer">The layer</a>
      <a href="#why">Why hybrid</a>
      <a href="#benchmarks">Benchmarks</a>
      <a href="#install">Install</a>
      <a href="#faq">FAQ</a>
      <a href="#contact">Contact</a>
    </nav>
    <button class="btn icon-btn" id="theme" aria-label="Switch between light and dark theme" title="Switch theme">
      <svg class="i-sun" viewBox="0 0 24 24" aria-hidden="true"><circle cx="12" cy="12" r="4.2"/><path d="M12 2.5v2M12 19.5v2M2.5 12h2M19.5 12h2M5.2 5.2l1.4 1.4M17.4 17.4l1.4 1.4M18.8 5.2l-1.4 1.4M6.6 17.4l-1.4 1.4"/></svg>
      <svg class="i-moon" viewBox="0 0 24 24" aria-hidden="true"><path d="M21 12.8A9 9 0 1 1 11.2 3a7 7 0 0 0 9.8 9.8z"/></svg>
    </button>
    <a class="btn btn-brand" href="https://github.com/AIAnytime/memrust">GitHub ↗</a>
  </div>
</header>

<main class="frame" id="top">

  <!-- HERO -->
  <section class="pad hero">
    <div>
      <p class="eyebrow">Memory infrastructure for AI agents</p>
      <h1>Your agent doesn't need a vector database.<br>It needs <span class="accent">memory</span>.</h1>
      <p class="lede" style="margin-top:1.6rem">
        A memory engine in Rust that answers <em>“what do I know about this?”</em> — not
        “which vectors are nearest?”
      </p>
      <div class="cta-row">
        <a class="btn btn-brand" href="#install">Install →</a>
        <a class="btn" href="https://github.com/AIAnytime/memrust">Read the source</a>
      </div>
      <div class="facts">
        <span>Rust</span><span>Apache-2.0</span><span>single binary</span><span>MCP built in</span><span>Python &amp; TypeScript SDKs</span>
      </div>
    </div>

    <div class="card" aria-label="An actual recall result from memrust">
      <div class="card-top">
        <span>memory.recall("what is connected to Project Phoenix")</span>
        <span>relational</span>
      </div>

      <div class="hit">
        <div class="hit-head"><span class="score tnum">0.0386</span><span class="kind">semantic</span><span style="color:var(--ink-3)">1m ago</span></div>
        <p class="hit-text">Dana Whitfield leads Project Phoenix</p>
        <div class="sigs">
          <div><div class="sig-name"><span>vector</span><span class="tnum">.0164</span></div><div class="track"><div class="fill v" style="width:100%"></div></div></div>
          <div><div class="sig-name"><span>lexical</span><span class="tnum">.0164</span></div><div class="track"><div class="fill l" style="width:100%;animation-delay:.06s"></div></div></div>
          <div><div class="sig-name"><span>graph</span><span class="tnum">.0164</span></div><div class="track"><div class="fill g" style="width:100%;animation-delay:.12s"></div></div></div>
          <div><div class="sig-name"><span>recency</span><span class="tnum">1.00</span></div><div class="track"><div class="fill r" style="width:100%;animation-delay:.18s"></div></div></div>
        </div>
      </div>

      <div class="hit">
        <div class="hit-head"><span class="score tnum">0.0309</span><span class="kind">semantic</span><span style="color:var(--ink-3)">1m ago</span></div>
        <p class="hit-text">the Billing Service rate limits at 100 requests per second</p>
        <div class="sigs">
          <div><div class="sig-name"><span>vector</span><span class="tnum">.0152</span></div><div class="track"><div class="fill v" style="width:93%;animation-delay:.24s"></div></div></div>
          <div><div class="sig-name"><span>lexical</span><span class="tnum">.0000</span></div><div class="track"><div class="fill l" style="width:0%"></div></div></div>
          <div><div class="sig-name"><span>graph</span><span class="tnum">.0159</span></div><div class="track"><div class="fill g" style="width:97%;animation-delay:.3s"></div></div></div>
          <div><div class="sig-name"><span>recency</span><span class="tnum">1.00</span></div><div class="track"><div class="fill r" style="width:100%;animation-delay:.36s"></div></div></div>
        </div>
      </div>

      <p class="card-note">
        That second memory never mentions Phoenix. It surfaced on <b>graph</b> — one hop
        through a shared entity — with lexical at zero.
      </p>
    </div>
  </section>


  <!-- THE MISSING LAYER -->
  <section class="seam pad" id="layer">
    <div class="layer-wrap">
      <div>
        <p class="eyebrow">The missing layer</p>
        <h2>Every other layer of the agent stack <span class="accent">already exists</span>.</h2>
        <p class="prose" style="margin-top:1.4rem">
          Models, orchestration, tools, evals — each has infrastructure built for it.
          Memory got a vector index and a shrug. So every team building agents writes the
          same layer by hand: a vector store, a keyword index, fusion logic, a graph for
          relationships, a scheduler for expiry and summarization, access control between
          agents, and a protocol shim to expose it all.
        </p>
        <p class="prose" style="margin-top:1rem">
          memrust is that layer as one binary. Not a database an agent queries —
          <strong>memory an agent has</strong>.
        </p>
      </div>

      <div class="stack" aria-label="Where memrust sits in the agent stack">
        <div class="tier">
          <span class="tier-l">Agent runtime</span>
          <span class="tier-v">Claude Code · LangGraph · OpenAI SDK · CrewAI</span>
        </div>
        <div class="tier tier-gap">
          <span class="tier-l">Tools &amp; orchestration</span>
          <span class="tier-v">MCP · function calling</span>
        </div>
        <div class="tier tier-me">
          <span class="tier-l">Memory</span>
          <span class="tier-v">remember · recall · forget</span>
          <span class="tier-badge">memrust</span>
        </div>
        <div class="tier">
          <span class="tier-l">Storage &amp; compute</span>
          <span class="tier-v">disk · embeddings · your infrastructure</span>
        </div>
      </div>
    </div>
  </section>

  <!-- THE 27% WALL -->
  <section class="seam pad wall" id="why">
    <div>
      <p class="eyebrow">The ceiling nobody talks about</p>
      <h2>Ask for <span class="accent">INC-90312</span> and vector search shrugs.</h2>
      <p class="prose" style="margin-top:1.4rem">
        Agents ask for exact things constantly: error codes, ticket IDs, commit SHAs,
        customer names, function names. Embeddings can't separate
        <span class="mono">INC-90312</span> from <span class="mono">INC-90319</span> —
        near-identical strings make near-identical vectors.
      </p>
      <p class="prose" style="margin-top:1rem">
        The telling part: <strong>exact brute-force vector search scores the same 27%.</strong>
        This isn't an index-quality problem a faster ANN fixes. It's the ceiling of the
        embedding itself, and every pure vector store inherits it.
      </p>
      <p class="prose" style="margin-top:1rem">
        memrust runs BM25 and an entity graph over the same memories and fuses the
        rankings. Same embeddings, same corpus — different architecture.
      </p>
    </div>

    <div>
      <div class="bars">
        <div class="bar-row muted">
          <div class="bar-label"><span>Exact vector search · the ceiling</span><b class="tnum">27%</b></div>
          <div class="bar-track"><div class="bar-fill" style="width:27%"></div></div>
        </div>
        <div class="bar-row muted">
          <div class="bar-label"><span>FAISS</span><b class="tnum">27%</b></div>
          <div class="bar-track"><div class="bar-fill" style="width:27%"></div></div>
        </div>
        <div class="bar-row muted">
          <div class="bar-label"><span>Chroma</span><b class="tnum">27%</b></div>
          <div class="bar-track"><div class="bar-fill" style="width:27%"></div></div>
        </div>
        <div class="bar-row muted">
          <div class="bar-label"><span>Qdrant</span><b class="tnum">27%</b></div>
          <div class="bar-track"><div class="bar-fill" style="width:27%"></div></div>
        </div>
        <div class="bar-row">
          <div class="bar-label"><span>memrust · hybrid</span><b class="tnum">75%</b></div>
          <div class="bar-track"><div class="bar-fill win" style="width:75%"></div></div>
        </div>
      </div>
      <p class="mono" style="margin-top:1rem;font-size:var(--step--1);color:var(--ink-3);line-height:1.6">
        hit@5 on 60 identifier lookups · 500 memories · identical all-MiniLM-L6-v2
        embeddings for every engine · benches/agent_recall.py
      </p>
    </div>
  </section>

  <!-- SIGNALS -->
  <section class="seam pad" id="signals">
    <p class="eyebrow">Four signals, one call</p>
    <h2 style="max-width:20ch">Retrieval that shows its <span class="accent">work</span>.</h2>
    <p class="prose" style="margin-top:1.3rem;margin-bottom:2.6rem">
      Every hit comes back with its score broken out by signal. An agent can tell a strong
      semantic match from a lucky keyword hit — and so can you, when something surfaces
      that shouldn't have.
    </p>
    <div class="grid-4">
      <div class="cell">
        <div class="cell-h"><span class="dot" style="background:var(--sig-vector)"></span>vector</div>
        <h3>Meaning</h3>
        <p>An in-crate HNSW index over your embeddings. Bring your own vectors, or let the engine call OpenAI, Gemini, or a local sentence-transformers server.</p>
      </div>
      <div class="cell">
        <div class="cell-h"><span class="dot" style="background:var(--sig-lexical)"></span>lexical</div>
        <h3>Exact terms</h3>
        <p>A BM25 inverted index for the identifiers, error codes and names that embeddings blur together.</p>
      </div>
      <div class="cell">
        <div class="cell-h"><span class="dot" style="background:var(--sig-graph)"></span>graph</div>
        <h3>Relationships</h3>
        <p>Entities extracted at ingest, linked by co-occurrence. Finds what's <em>connected</em> to a thing, not just what resembles it. No external graph database.</p>
      </div>
      <div class="cell">
        <div class="cell-h"><span class="dot" style="background:var(--sig-recency)"></span>recency</div>
        <h3>Time</h3>
        <p>Exponential decay as a first-class ranking term, because "what did we decide recently" is not a similarity question.</p>
      </div>
    </div>
  </section>

  <!-- STATS -->
  <section class="seam">
    <div class="pad" style="padding-bottom:0">
      <p class="eyebrow">Measured, not claimed</p>
      <h2 style="max-width:24ch">Every number here comes from a script in the repo.</h2>
    </div>
    <div class="stats" style="margin-top:2.6rem">
      <div class="stat">
        <div class="stat-v tnum">0.64<small>ms</small></div>
        <div class="stat-l">recall p50 at 20k vectors — second-fastest server measured, behind pgvector</div>
      </div>
      <div class="stat">
        <div class="stat-v tnum">1.000</div>
        <div class="stat-l">recall@10 against exact brute-force ground truth</div>
      </div>
      <div class="stat">
        <div class="stat-v tnum">4<small>×</small></div>
        <div class="stat-l">less vector memory with SQ8, applied automatically at 1024+ dims where it also runs faster</div>
      </div>
      <div class="stat">
        <div class="stat-v tnum">1,000</div>
        <div class="stat-l">acknowledged writes, 1,000 recovered after <span class="mono">kill -9</span> mid-checkpoint</div>
      </div>
    </div>
  </section>

  <!-- COMPARISON -->
  <section class="seam pad" id="benchmarks">
    <p class="eyebrow">What you'd otherwise assemble</p>
    <h2 style="max-width:26ch">One engine instead of <span class="accent">five systems</span>.</h2>
    <p class="prose" style="margin-top:1.3rem;margin-bottom:2.2rem">
      The usual answer to agent memory is a vector DB, a keyword index, fusion code, a graph
      store, a scheduler for expiry and summarization, access control, and an MCP shim.
      That's the stack memrust replaces.
    </p>
    <div class="scroller">
      <table>
        <thead>
          <tr><th>Capability</th><th>memrust</th><th>Qdrant</th><th>Chroma</th><th>FAISS</th><th>pgvector</th><th>LanceDB</th></tr>
        </thead>
        <tbody>
          <tr><td>Vector search</td><td class="yes"></td><td class="yes"></td><td class="yes"></td><td class="yes"></td><td class="yes"></td><td class="yes"></td></tr>
          <tr><td>Keyword fused into one query</td><td class="yes"></td><td class="partial">setup</td><td class="no"></td><td class="no"></td><td class="partial">DIY</td><td class="yes"></td></tr>
          <tr><td>Entity graph as a signal</td><td class="yes"></td><td class="no"></td><td class="no"></td><td class="no"></td><td class="partial">DIY</td><td class="no"></td></tr>
          <tr><td>Per-signal score explanation</td><td class="yes"></td><td class="no"></td><td class="no"></td><td class="no"></td><td class="no"></td><td class="no"></td></tr>
          <tr><td>Agent API + memory kinds</td><td class="yes"></td><td class="no"></td><td class="no"></td><td class="no"></td><td class="no"></td><td class="no"></td></tr>
          <tr><td>TTLs &amp; automatic consolidation</td><td class="yes"></td><td class="no"></td><td class="no"></td><td class="no"></td><td class="partial">cron</td><td class="no"></td></tr>
          <tr><td>Multi-agent private/shared memory</td><td class="yes"></td><td class="partial">filters</td><td class="partial">filters</td><td class="no"></td><td class="partial">policies</td><td class="partial">filters</td></tr>
          <tr><td>MCP server for agent runtimes</td><td class="yes"></td><td class="no"></td><td class="no"></td><td class="no"></td><td class="no"></td><td class="no"></td></tr>
          <tr><td>Built-in dashboard</td><td class="yes"></td><td class="yes"></td><td class="no"></td><td class="no"></td><td class="no"></td><td class="no"></td></tr>
        </tbody>
      </table>
    </div>
    <p class="mono" style="margin-top:1rem;font-size:var(--step--1);color:var(--ink-3)">
      ✓ built in · setup/DIY/cron/filters/policies = possible, you build it · — not available
    </p>

    <h3 style="margin-top:3.2rem">Vector search, head to head</h3>
    <p class="prose" style="margin-top:.8rem;margin-bottom:1.4rem">
      20,000 vectors, 384 dims, identical HNSW settings, recall against exact ground truth.
      Query latency is second-best among the servers and recall ties the field. Ingest is
      slower on purpose: memrust fsyncs every write, where the others buffer. For a memory
      layer that's the right trade — an agent quietly losing what it was told is worse than
      writing a little slower.
    </p>
    <div class="scroller">
      <table>
        <thead><tr><th>Engine</th><th>Ingest/s</th><th>Query p50</th><th>recall@10</th></tr></thead>
        <tbody>
          <tr><td>FAISS <span class="mono" style="color:var(--ink-3)">in-process</span></td><td class="num">35,625</td><td class="num">0.06 ms</td><td class="num">1.000</td></tr>
          <tr><td>LanceDB <span class="mono" style="color:var(--ink-3)">embedded</span></td><td class="num">99,238</td><td class="num">8.60 ms</td><td class="num">1.000</td></tr>
          <tr><td>Chroma <span class="mono" style="color:var(--ink-3)">in-process</span></td><td class="num">5,907</td><td class="num">0.49 ms</td><td class="num">1.000</td></tr>
          <tr><td>pgvector <span class="mono" style="color:var(--ink-3)">server</span></td><td class="num">5,017</td><td class="num">0.46 ms</td><td class="num">1.000</td></tr>
          <tr><td>Qdrant <span class="mono" style="color:var(--ink-3)">server</span></td><td class="num">2,475</td><td class="num">2.08 ms</td><td class="num">1.000</td></tr>
          <tr class="me"><td>memrust <span class="mono" style="color:var(--ink-3)">server</span></td><td class="num">988</td><td class="num">0.64 ms</td><td class="num">1.000</td></tr>
        </tbody>
      </table>
    </div>
  </section>

  <!-- INSTALL -->
  <section class="seam pad" id="install">
    <p class="eyebrow">Two minutes to first recall</p>
    <h2 style="max-width:22ch">Install it, <span class="accent">remember</span> something, ask.</h2>

    <div style="margin-top:2.4rem">
      <div class="tabs" role="tablist" aria-label="Choose a language">
        <button class="tab" role="tab" aria-selected="true" aria-controls="p-sh" id="t-sh">shell</button>
        <button class="tab" role="tab" aria-selected="false" aria-controls="p-py" id="t-py">python</button>
        <button class="tab" role="tab" aria-selected="false" aria-controls="p-ts" id="t-ts">typescript</button>
        <button class="tab" role="tab" aria-selected="false" aria-controls="p-rs" id="t-rs">rust</button>
        <button class="tab" role="tab" aria-selected="false" aria-controls="p-mcp" id="t-mcp">mcp</button>
      </div>

      <div class="card">
        <div role="tabpanel" id="p-sh" aria-labelledby="t-sh"><pre><span class="c"># the engine — dashboard and HTTP API on :7700</span>
cargo install memrust
memrust serve

<span class="c"># or with Docker</span>
docker run -p 7700:7700 -v memrust-data:/data memrust</pre></div>

        <div role="tabpanel" id="p-py" aria-labelledby="t-py" hidden><pre><span class="c"># pip install memrust</span>
<span class="k">from</span> memrust <span class="k">import</span> MemrustClient

memory = <span class="f">MemrustClient</span>(<span class="s">"http://127.0.0.1:7700"</span>, agent_id=<span class="s">"planner"</span>)
memory.<span class="f">remember</span>(<span class="s">"Deploy 47 failed with error E1234"</span>, kind=<span class="s">"episodic"</span>)

<span class="k">for</span> hit <span class="k">in</span> memory.<span class="f">recall</span>(<span class="s">"what went wrong with the deploy?"</span>):
    <span class="f">print</span>(hit[<span class="s">"score"</span>], hit[<span class="s">"record"</span>][<span class="s">"text"</span>], hit[<span class="s">"signals"</span>])</pre></div>

        <div role="tabpanel" id="p-ts" aria-labelledby="t-ts" hidden><pre><span class="c">// npm install memrust-client</span>
<span class="k">import</span> { MemrustClient } <span class="k">from</span> <span class="s">"memrust-client"</span>;

<span class="k">const</span> memory = <span class="k">new</span> <span class="f">MemrustClient</span>(<span class="s">"http://127.0.0.1:7700"</span>, { agentId: <span class="s">"planner"</span> });
<span class="k">await</span> memory.<span class="f">remember</span>(<span class="s">"the Billing Service rate limits at 100 rps"</span>);

<span class="k">const</span> hits = <span class="k">await</span> memory.<span class="f">recall</span>(<span class="s">"Project Phoenix"</span>, { strategy: <span class="s">"relational"</span> });</pre></div>

        <div role="tabpanel" id="p-rs" aria-labelledby="t-rs" hidden><pre><span class="c">// cargo add memrust</span>
<span class="k">use</span> memrust::{MemoryEngine, RememberRequest, RecallRequest};

<span class="k">let mut</span> memory = MemoryEngine::<span class="f">open</span>(<span class="s">"./data"</span>.as_ref())?;
memory.<span class="f">remember</span>(RememberRequest {
    text: <span class="s">"user prefers concise answers"</span>.into(),
    ..Default::<span class="f">default</span>()
})?;

<span class="k">let</span> hits = memory.<span class="f">recall</span>(&RecallRequest {
    query: <span class="s">"what does the user prefer?"</span>.into(),
    ..Default::<span class="f">default</span>()
});</pre></div>

        <div role="tabpanel" id="p-mcp" aria-labelledby="t-mcp" hidden><pre><span class="c"># give Claude Code persistent memory</span>
claude mcp add memory -- memrust mcp --data-dir ~/.memrust --agent-id planner

<span class="c"># the agent gets three tools:</span>
<span class="c">#   memory_remember · memory_recall · memory_forget</span>
<span class="c"># recall returns the per-signal breakdown, so the agent can</span>
<span class="c"># reason about why something surfaced.</span></pre></div>
      </div>
    </div>

    <div class="facts" style="margin-top:1.6rem">
      <span>crates.io</span><span>PyPI</span><span>npm</span><span>prebuilt Linux &amp; macOS binaries</span>
    </div>
  </section>

  <!-- WHERE IT'S GOING -->
  <section class="seam pad">
    <p class="eyebrow">Where this is going</p>
    <h2 style="max-width:24ch">The engine is open. The <span class="accent">hard part</span> is shared memory.</h2>
    <p class="prose" style="margin-top:1.3rem">
      memrust is Apache-2.0 and will stay that way — an engine you can read, fork and run
      on your own hardware. What comes next is the part single-node can't solve: fleets of
      agents sharing memory across machines, with isolation, quotas and replication. Open
      core, hosted control plane.
    </p>
    <div class="road">
      <div class="road-item"><span class="road-t">Shipping now</span><p>Hybrid retrieval, memory lifecycle, multi-agent visibility, MCP, dashboard, SDKs.</p></div>
      <div class="road-item"><span class="road-t">Next</span><p>Streaming replication, per-namespace quotas, LLM entity extraction, hierarchical summarization.</p></div>
      <div class="road-item"><span class="road-t">Then</span><p>Hosted memory with per-agent isolation — the engine stays open, the control plane is the product.</p></div>
    </div>
  </section>


  <!-- FAQ -->
  <section class="seam pad" id="faq">
    <p class="eyebrow">Straight answers</p>
    <h2 style="max-width:22ch">The questions we <span class="accent">actually get</span>.</h2>

    <div class="faq" style="margin-top:2.4rem">
      <details open>
        <summary>Is this just Qdrant + Mem0 in one box?</summary>
        <div class="a">
          <p>That stack is two systems: the memory semantics live in Python, the storage lives
          somewhere else, they talk over a network, and an LLM sits in the loop deciding what
          to keep. memrust implements the memory semantics <em>inside</em> the storage engine.</p>
          <p>That's why filters apply during index traversal instead of after it, why it runs
          with no API keys at all, and why durability is something you can reason about rather
          than inherit from whatever store you configured underneath.</p>
        </div>
      </details>

      <details>
        <summary>So is memrust a vector database?</summary>
        <div class="a">
          <p>There's one inside it — HNSW, scalar quantization, a write-ahead log, checkpoint
          recovery. But that index is one of four retrieval signals, and it isn't the API.</p>
          <p>A vector database stores an embedding with a metadata blob attached. memrust
          stores a <em>memory</em>: typed as episodic, semantic, working or procedural; owned
          by an agent; able to expire; linked to extracted entities; and carrying provenance
          when it was distilled from older memories.</p>
        </div>
      </details>

      <details>
        <summary>Do I need an LLM or API keys?</summary>
        <div class="a">
          <p>No. The default embedder and summarizer run offline with zero keys — enough to
          try every feature on this page, including the notebooks.</p>
          <p>For production quality, point it at OpenAI, Gemini, or a local
          sentence-transformers server through Ollama, TEI, LM Studio or vLLM. Or skip the
          embedder entirely and pass your own vectors.</p>
        </div>
      </details>

      <details>
        <summary>Why not just bolt a keyword index onto my vector DB?</summary>
        <div class="a">
          <p>You can — and if you do, you'll rebuild what memrust already is: two indexes,
          rank fusion, a graph for relationships, jobs for expiry and consolidation, per-agent
          access rules, and a protocol shim so agents can reach it.</p>
          <p>The parts that resist bolting on are pre-filtering <em>inside</em> index
          traversal and per-signal score explanation. Both have to live in the engine; neither
          can be added by a layer sitting above it.</p>
        </div>
      </details>

      <details>
        <summary>Can I keep the embedding model I already use?</summary>
        <div class="a">
          <p>Yes, and it's the common path. Pass <span class="mono">embedding</span> on write
          and <span class="mono">query_embedding</span> on recall, and memrust never calls a
          model — BGE, E5, OpenAI, Voyage, whatever you already run.</p>
          <p>The first vector fixes a collection's dimension. Mixing models afterwards is
          rejected with a clear error rather than silently returning nonsense similarities.</p>
        </div>
      </details>

      <details>
        <summary>Is it production-ready?</summary>
        <div class="a">
          <p>It's young and single-node, so be deliberate. What's there: API keys scoped to
          namespaces, isolated stores with their own indexes and directories, and durability
          tested under <span class="mono">kill -9</span> with concurrent writers and
          checkpoints racing — 1,000 acknowledged writes, 1,000 recovered — plus recall@10 of
          1.000 against exact ground truth and a suite that runs on every commit.</p>
          <p>What it isn't: distributed, multi-writer, or scarred by years in production. Run
          it where a single node fits, which covers most agent workloads today.</p>
        </div>
      </details>

      <details>
        <summary>How does it fit into my agent framework?</summary>
        <div class="a">
          <p>Three ways. As an <span class="mono">MCP</span> server, which is how Claude Code
          and Claude Desktop pick it up — the agent gets remember, recall and forget as native
          tools. Over HTTP with the Python or TypeScript SDK, for LangGraph, CrewAI, the
          OpenAI SDK or anything else. Or as a Rust library, embedded with no server at all.</p>
        </div>
      </details>

      <details>
        <summary>What's the license, and how does this make money?</summary>
        <div class="a">
          <p>Apache-2.0, permanently. The engine is meant to be read, forked and self-hosted,
          and nothing on this page is behind a paywall.</p>
          <p>The commercial layer is the part a single node can't do: fleets of agents sharing
          memory across machines, with isolation, quotas and replication. Open core, hosted
          control plane.</p>
        </div>
      </details>

      <details>
        <summary>Why Rust?</summary>
        <div class="a">
          <p>Because this is a storage engine. Cache-friendly layouts, SIMD-friendly distance
          kernels, and predictable latency with no garbage collector pausing in the middle of
          a recall.</p>
          <p>It's also why the whole thing ships as one static binary with no runtime to
          install — the same reason Qdrant is written in Rust.</p>
        </div>
      </details>
    </div>
  </section>

  <!-- CONTACT -->
  <section class="seam pad" id="contact">
    <div class="contact-wrap">
      <div>
        <p class="eyebrow">Get in touch</p>
        <h2>Building something that needs to <span class="accent">remember</span>?</h2>
        <p class="prose" style="margin-top:1.3rem">
          Whether you're wiring memrust into an agent today, hitting an edge we haven't,
          or looking at the memory layer as a market — the fastest way to reach us is
          directly.
        </p>
        <div class="cta-row">
          <a class="btn btn-brand" href="mailto:sroy27.ai@gmail.com?subject=memrust">Email us →</a>
          <a class="btn" href="https://github.com/AIAnytime/memrust">Star on GitHub ↗</a>
        </div>
      </div>
      <div class="contact-cards">
        <a class="contact-card" href="https://github.com/AIAnytime/memrust/issues">
          <span class="cc-k">Bugs &amp; questions</span>
          <span class="cc-v">GitHub Issues</span>
          <p>Fastest for anything reproducible. Benchmarks and tests are in the repo.</p>
        </a>
        <a class="contact-card" href="mailto:sroy27.ai@gmail.com?subject=memrust%20in%20production">
          <span class="cc-k">Using it in production</span>
          <span class="cc-v">Email</span>
          <p>Tell us the workload. Scale, latency budget and agent count shape what we build next.</p>
        </a>
        <a class="contact-card" href="mailto:sroy27.ai@gmail.com?subject=memrust%20—%20investment%20%26%20partnerships">
          <span class="cc-k">Investors &amp; partners</span>
          <span class="cc-v">Email</span>
          <p>Happy to walk through the architecture, the benchmarks and where the hosted layer goes.</p>
        </a>
      </div>
    </div>
  </section>

  <footer class="seam pad">
    <div class="foot">
      <span>memrust · Apache-2.0 · built by <a href="https://github.com/AIAnytime">AI Anytime</a></span>
      <span class="foot-links">
        <a href="https://github.com/AIAnytime/memrust">GitHub</a>
        <a href="https://crates.io/crates/memrust">crates.io</a>
        <a href="https://pypi.org/project/memrust/">PyPI</a>
        <a href="https://www.npmjs.com/package/memrust-client">npm</a>
        <a href="https://github.com/AIAnytime/memrust/tree/main/benches">Benchmarks</a>
        <a href="mailto:sroy27.ai@gmail.com?subject=memrust">Contact</a>
      </span>
    </div>
  </footer>
</main>

<script>
  // Theme: follow the OS, remember an explicit choice.
  var root = document.documentElement;
  document.getElementById("theme").addEventListener("click", function () {
    var set = root.getAttribute("data-theme");
    var dark = set ? set === "dark" : matchMedia("(prefers-color-scheme: dark)").matches;
    var next = dark ? "light" : "dark";
    root.setAttribute("data-theme", next);
    try { localStorage.setItem("memrust-site-theme", next); } catch (e) {}
  });

  // Language tabs.
  var tabs = Array.prototype.slice.call(document.querySelectorAll(".tab"));
  tabs.forEach(function (tab) {
    tab.addEventListener("click", function () {
      tabs.forEach(function (t) {
        var on = t === tab;
        t.setAttribute("aria-selected", String(on));
        document.getElementById(t.getAttribute("aria-controls")).hidden = !on;
      });
    });
  });
</script>
</body>
</html>