Repository navigation
Expand file tree
/
Copy pathindex.html
More file actions
772 lines (710 loc) · 38 KB
/
Copy pathindex.html
File metadata and controls
772 lines (710 loc) · 38 KB
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
<!DOCTYPE html>
<html lang="en">
<head>
<!-- Google Consent Mode v2 & GA4 (DSGVO / GDPR Compliant) -->
<script>
window.dataLayer = window.dataLayer || [];
function gtag(){dataLayer.push(arguments);}
// Initialize default consent from domain cookie or storage fallback
var storedConsent = null;
try {
var match = document.cookie.match(/(^|;)\s*okf_cookie_consent\s*=\s*([^;]+)/);
if (match) {
storedConsent = decodeURIComponent(match[2]);
} else {
storedConsent = localStorage.getItem('okf_cookie_consent');
}
} catch (e) {}
gtag('consent', 'default', {
'analytics_storage': storedConsent === 'granted' ? 'granted' : 'denied',
'ad_storage': 'denied',
'ad_user_data': 'denied',
'ad_personalization': 'denied',
'wait_for_update': 500
});
gtag('js', new Date());
gtag('config', 'G-CCKX1QN84C', {
'anonymize_ip': true
});
</script>
<script async src="https://www.googletagmanager.com/gtag/js?id=G-CCKX1QN84C"></script>
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>OKF Agent Memory — Git-Native Persistent Memory for AI Coding Agents</title>
<meta name="description"
content="Sub-300µs, zero-dependency in-memory BM25 search, embedded MCP server, and progressive disclosure for Claude Code, Cursor, Codex, and local LLMs. Standardized on Google OKF v0.2.">
<link rel="canonical" href="https://okf-memory.dev/">
<link rel="icon" type="image/svg+xml"
href="data:image/svg+xml,<svg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 100 100'><text y='.9em' font-size='90'>🧠</text></svg>">
<!-- Open Graph / Facebook -->
<meta property="og:type" content="website">
<meta property="og:url" content="https://okf-memory.dev/">
<meta property="og:title" content="OKF Agent Memory — Git-Native Persistent Memory for AI Coding Agents">
<meta property="og:description"
content="Sub-300µs, zero-dependency in-memory BM25 search, embedded MCP server, and progressive disclosure for Claude Code, Cursor, and Local LLMs. Standardized on Google OKF v0.2.">
<meta property="og:image" content="https://okf-memory.dev/assets/og-preview.png">
<!-- Twitter -->
<meta name="twitter:card" content="summary_large_image">
<meta name="twitter:url" content="https://okf-memory.dev/">
<meta name="twitter:title" content="OKF Agent Memory — Git-Native Persistent Memory for AI Coding Agents">
<meta name="twitter:description"
content="Sub-300µs, zero-dependency in-memory BM25 search, embedded MCP server, and progressive disclosure for Claude Code, Cursor, and Local LLMs. Standardized on Google OKF v0.2.">
<meta name="twitter:image" content="https://okf-memory.dev/assets/og-preview.png">
<link rel="preconnect" href="https://fonts.googleapis.com">
<link rel="preconnect" href="https://fonts.gstatic.com" crossorigin>
<link
href="https://fonts.googleapis.com/css2?family=JetBrains+Mono:wght@400;500;700&family=Outfit:wght@400;500;600;700;800&display=swap"
rel="stylesheet">
<link rel="stylesheet" href="style.css">
</head>
<body>
<!-- Navigation -->
<header class="site-header">
<div class="container nav-wrap">
<a href="#" class="brand-logo">
<div class="logo-icon">🧠</div>
<span>OKF Agent Memory</span>
</a>
<nav>
<ul class="nav-links">
<li><a href="#features">Features</a></li>
<li><a href="#benchmarks">Benchmarks</a></li>
<li><a href="#architecture">Architecture</a></li>
<li><a href="#quickstart">Quickstart</a></li>
<li><a href="https://registry.okf-memory.dev" target="_blank" rel="noopener">Registry</a></li>
<li><a href="#faq">FAQ</a></li>
<li><a href="https://github.com/okf-memory/okf-agent-memory" target="_blank" rel="noopener">GitHub</a></li>
</ul>
</nav>
<div class="nav-actions">
<a href="#waitlist" class="btn btn-primary">Join Cloud Beta</a>
</div>
</div>
</header>
<main>
<!-- Hero Section -->
<section class="hero-section">
<div class="container">
<div class="badge-wrapper">
<span class="badge-pulse"></span>
<span>Release v0.6.1 • Sub-300µs Search • <a href="https://trendshift.io/repositories/215663" target="_blank" rel="noopener noreferrer" class="badge-link">🔥 #2 Go Trending on Trendshift</a></span>
</div>
<h1 class="hero-title">
The Git-Native Memory for <br><span class="hero-highlight">Autonomous AI Coding Agents</span>
</h1>
<p class="hero-subtitle">
Transparent, auditable Markdown memory for Claude Code, Cursor, and Local LLMs. Slash token overhead by 80%
through in-memory BM25 progressive disclosure — with zero external database dependencies.
</p>
<div class="hero-cta-group">
<div class="install-switcher">
<div class="install-tabs">
<button class="install-tab-btn active" data-install="curl">cURL 1-Liner</button>
<button class="install-tab-btn" data-install="go">Go Install</button>
<button class="install-tab-btn" data-install="source">Source Build</button>
<button class="install-tab-btn" data-install="brew">Homebrew</button>
</div>
<div id="hero-install-pill" class="install-pill" data-copy="curl -fsSL https://okf-memory.dev/install.sh | sh">
<span id="hero-install-cmd">$ curl -fsSL https://okf-memory.dev/install.sh | sh</span>
<span class="copy-icon">📋</span>
</div>
</div>
<a href="#quickstart" class="btn btn-secondary">Get Started in 1 Min</a>
<a href="https://registry.okf-memory.dev" class="btn btn-secondary" target="_blank" rel="noopener">Browse Registry 🗂️</a>
<a href="#waitlist" class="btn btn-glow">Get OKF Cloud Access</a>
</div>
<!-- Interactive Terminal Simulator -->
<div class="terminal-window">
<div class="terminal-header">
<div class="terminal-dots">
<span class="dot dot-red"></span>
<span class="dot dot-yellow"></span>
<span class="dot dot-green"></span>
</div>
<div class="terminal-tabs">
<button class="term-tab btn-secondary btn active" style="padding: 4px 10px; font-size: 0.75rem;"
data-cmd="search">okf search</button>
<button class="term-tab btn-secondary btn" style="padding: 4px 10px; font-size: 0.75rem;"
data-cmd="bootstrap">okf bootstrap</button>
<button class="term-tab btn-secondary btn" style="padding: 4px 10px; font-size: 0.75rem;"
data-cmd="validate">okf validate</button>
<button class="term-tab btn-secondary btn" style="padding: 4px 10px; font-size: 0.75rem;"
data-cmd="concept">📄 auth-decision.md</button>
</div>
<div class="terminal-title">zsh — okf-agent-memory</div>
</div>
<div class="terminal-body">
<pre id="terminal-code-output"><span class="term-cmd">$ okf search "jwt auth flow" knowledge</span>
Found 1 matching concept(s) in 'knowledge':
1. <span class="term-purple">[4.82]</span> <span class="term-green">architecture/auth-decision</span> (<span class="term-yellow">Decision</span>)
<span class="term-dim">Standardized RSA-256 JWTs with 15m expiration & sliding refresh tokens.</span>
Matches: <span class="term-green">title, description, tags</span>
<span class="term-stat">⚡ In-Memory BM25 search completed in 268.4µs (<0.3ms)</span></pre>
</div>
</div>
</div>
</section>
<!-- Key Features Grid -->
<section id="features" class="section">
<div class="container">
<div class="section-header">
<span class="section-tag">Core Value</span>
<h2 class="section-title">Engineered for Maximum Agent Performance</h2>
<p class="section-desc">Why developers and enterprises choose OKF over opaque vector databases and
unstructured flat files.</p>
</div>
<div class="features-grid">
<div class="feature-card">
<div class="feature-icon">⚡</div>
<h3 class="feature-title">Sub-300µs BM25 Search</h3>
<p class="feature-text">Built in pure Go with zero runtime dependencies. Queries full knowledge graphs in
microseconds directly in memory without slow vector embeddings or external API calls.</p>
</div>
<div class="feature-card">
<div class="feature-icon">🗂️</div>
<h3 class="feature-title">Progressive Disclosure</h3>
<p class="feature-text">Slash token costs by up to 80%. Instead of dumping 20k-token monoliths into every
prompt, agents dynamically pull only the exact 300-token concept required.</p>
</div>
<div class="feature-card">
<div class="feature-icon">🔒</div>
<h3 class="feature-title">Trust & Provenance Tiers</h3>
<p class="feature-text">Strict distinction between authoritative human decisions
(<code>verified: human</code>) and agent drafts (<code>generated: agent</code>). Agents never overwrite
human architectural law.</p>
</div>
<div class="feature-card">
<div class="feature-icon">🔌</div>
<h3 class="feature-title">Built-In Stdio MCP Server</h3>
<p class="feature-text">Plug-and-play Model Context Protocol (MCP) server embedded inside the single binary.
Works seamlessly with Claude Code, Cursor, Windsurf, and custom IDEs.</p>
</div>
<div class="feature-card">
<div class="feature-icon">🛡️</div>
<h3 class="feature-title">100% Git-Native & Auditable</h3>
<p class="feature-text">No proprietary database silos. Your memory lives in human-readable Markdown files
right in your repository. Use <code>git diff</code>, <code>git blame</code>, and code reviews.</p>
</div>
<div class="feature-card">
<div class="feature-icon">🚀</div>
<h3 class="feature-title">1-Command Bootstrapping</h3>
<p class="feature-text">Run <code>okf bootstrap .</code> in any new or existing repository. Automatically
scaffolds knowledge bundles, non-destructively enriches <code>AGENTS.md</code>, and installs skills.</p>
</div>
</div>
<!-- Local LLM & Privacy Callout -->
<div class="local-ai-banner">
<div class="local-ai-icon">⚡</div>
<div class="local-ai-content">
<h4>
Up to 5.2x Faster TTFT for Local LLMs (Ollama / LM Studio / Qwen / Mistral)
<span class="local-ai-tag">100% OFFLINE</span>
</h4>
<p>
Eliminates the local AI bottleneck: Feeding 20k-token monolith files into local 8B–26B models causes
severe prompt prefill latency (18s+ on Metal GPU) and exhausts VRAM KV-caches. With OKF Progressive Disclosure, agents
query in <300µs and ingest only the exact 300-token concept required — speeding up TTFT by up to 5.2x
without hallucinations.
</p>
</div>
</div>
</div>
</section>
<!-- Benchmark Comparison & Empirical LLM Results -->
<section id="benchmarks" class="section">
<div class="container">
<div class="section-header">
<span class="section-tag">Live Empirical Benchmarks</span>
<h2 class="section-title">Tested Across Local & Remote Models</h2>
<p class="section-desc">
Direct empirical measurements proving OKF's dual impact: slashing <strong>80.1% prompt token billing</strong> on <strong>Remote Cloud APIs</strong> and eliminating <strong>brutal prefill latency (up to 5.2x faster TTFT)</strong> on <strong>Local On-Device Models</strong>.
</p>
</div>
<!-- Hardware / Cloud Environment Banner -->
<div class="benchmark-hardware-card">
<div class="hw-icon-pulse" id="hw-icon">☁️</div>
<div class="hw-content">
<div class="hw-badge-row">
<span class="hw-pill-live hw-pill-cloud" id="hw-pill"><span class="hw-dot cloud-dot" id="hw-dot"></span><span id="hw-pill-text">REMOTE MODEL • Official OpenAI Cloud API</span></span>
<span class="hw-date">Test Date: September 2026</span>
</div>
<h3 class="hw-machine-title" id="hw-machine-title">OpenAI Production Cloud • api.openai.com</h3>
<p class="hw-machine-details" id="hw-machine-details">
<strong>Deployment:</strong> Remote Cloud SaaS • <strong>Model:</strong> <code>gpt-5.6-sol</code> •
<strong>Primary Benefit:</strong> 2.4x Faster TTFT & -75.7% Prompt Context Overhead (3,771 → 917 tokens)
</p>
</div>
<div class="hw-actions">
<a id="hw-report-link" href="https://github.com/okf-memory/okf-agent-memory/blob/main/benchmarks/results/BENCHMARK_RESULTS_DMAA_openai_gpt-5.6-sol.md" target="_blank" rel="noopener" class="btn btn-secondary btn-sm">
View Raw Report ↗
</a>
</div>
</div>
<!-- Categorized Model Switcher: Remote vs Local -->
<div class="model-switcher-wrap">
<div class="model-switcher-groups">
<!-- Group 1: Remote Cloud API -->
<div class="model-category-block block-cloud">
<div class="model-category-header">
<span class="cat-pill cat-pill-cloud">☁️ Remote / Cloud API</span>
<span class="cat-sub">Pay-per-Token SaaS • Billing & Attention Focus</span>
</div>
<div class="model-category-tabs">
<button class="model-tab-btn model-tab-remote active" data-model="openai">
<span class="tab-chip">OpenAI Flagship</span>
<span class="tab-name">GPT-5.6 Sol</span>
</button>
<button class="model-tab-btn model-tab-remote" data-model="claude">
<span class="tab-chip">Anthropic Frontier</span>
<span class="tab-name">Claude Opus 5</span>
</button>
</div>
</div>
<!-- Group 2: Local On-Device -->
<div class="model-category-block block-local">
<div class="model-category-header">
<span class="cat-pill cat-pill-local">🖥️ Local / On-Device</span>
<span class="cat-sub">Metal GPU • TTFT Acceleration Focus</span>
</div>
<div class="model-category-tabs">
<button class="model-tab-btn model-tab-local" data-model="qwen14b">
<span class="tab-chip">14B Code Flagship</span>
<span class="tab-name">Qwen 2.5 Coder 14B</span>
</button>
<button class="model-tab-btn model-tab-local" data-model="gemma26b">
<span class="tab-chip">26B MoE</span>
<span class="tab-name">Google Gemma 4 26B</span>
</button>
<button class="model-tab-btn model-tab-local" data-model="qwen30b">
<span class="tab-chip">30B Heavyweight</span>
<span class="tab-name">Qwen 3 Coder 30B</span>
</button>
</div>
</div>
</div>
</div>
<!-- Live Highlight Metric Cards -->
<div class="benchmark-metrics-grid">
<div class="b-card b-card-glow">
<div class="b-card-header">
<span class="b-card-label">Context Overhead Reduction</span>
<span class="b-delta-badge" id="metric-delta-tokens">-75.7%</span>
</div>
<div class="b-card-value" id="metric-tokens-val">917 <span class="unit">tokens</span></div>
<div class="b-card-sub" id="metric-tokens-sub">vs. 3,771 tokens in Monolith Dump</div>
<div class="b-progress-bar">
<div class="b-progress-fill tokens-fill" style="width: 24.3%;"></div>
</div>
</div>
<div class="b-card b-card-glow">
<div class="b-card-header">
<span class="b-card-label">Prefill Latency (TTFT)</span>
<span class="b-delta-badge badge-green" id="metric-delta-ttft">2.4x Faster</span>
</div>
<div class="b-card-value" id="metric-ttft-val">11,565 <span class="unit">ms</span></div>
<div class="b-card-sub" id="metric-ttft-sub">vs. 27,213 ms cloud TTFT latency</div>
<div class="b-progress-bar">
<div class="b-progress-fill ttft-fill" style="width: 42.5%;"></div>
</div>
</div>
<div class="b-card">
<div class="b-card-header">
<span class="b-card-label">Total Turn Duration</span>
<span class="b-delta-badge badge-cyan" id="metric-delta-time">-41.9s</span>
</div>
<div class="b-card-value" id="metric-time-val">43.0 <span class="unit">sec</span></div>
<div class="b-card-sub" id="metric-time-sub">vs. 84.9 sec full monolith turn (2.0x faster)</div>
<div class="b-progress-bar">
<div class="b-progress-fill time-fill" style="width: 50.7%;"></div>
</div>
</div>
<div class="b-card">
<div class="b-card-header">
<span class="b-card-label">Policy Adherence Accuracy</span>
<span class="b-delta-badge badge-green" id="metric-delta-compliance">100% PASS</span>
</div>
<div class="b-card-value" id="metric-compliance-val">9/9 <span class="unit">checks</span></div>
<div class="b-card-sub" id="metric-compliance-sub">AES-256-GCM • 96-bit Nonce • AAG Directives • Provenance</div>
<div class="b-compliance-pills">
<span class="c-pill">✓ GCM</span>
<span class="c-pill">✓ 12B Nonce</span>
<span class="c-pill">✓ Version Header</span>
<span class="c-pill">✓ AAG Directives</span>
</div>
</div>
</div>
<!-- Head-to-Head Comparison Panels -->
<div class="head-to-head-grid">
<!-- Monolith Approach -->
<div class="h2h-panel monolith-panel">
<div class="h2h-header">
<div class="h2h-title-wrap">
<span class="h2h-pill-danger">Traditional Pattern</span>
<h4>Monolith Context Dump (CLAUDE.md)</h4>
</div>
<span class="h2h-score" id="h2h-mono-tokens">3,771 Tokens</span>
</div>
<p class="h2h-desc">
Every turn dumps the entire 11.5 KB documentation corpus into the system prompt. Forces the model to parse megabytes of unrelated rules.
</p>
<ul class="h2h-list">
<li><span class="cross">✕</span> <strong>Severe Prefill Lag:</strong> <span id="h2h-mono-ttft">27,213 ms</span> time-to-first-token</li>
<li><span class="cross">✕</span> <strong>O(N) Token Bloat:</strong> Scales linearly with team documentation size</li>
<li><span class="cross">✕</span> <strong>Attention Dispersion:</strong> Critical ADR policies get lost in context walls</li>
<li><span class="cross">✕</span> <strong>Cloud Cost Escalation:</strong> Wastes 75%+ of budget on repeat prompt processing</li>
</ul>
</div>
<!-- OKF Progressive Disclosure -->
<div class="h2h-panel okf-panel">
<div class="h2h-header">
<div class="h2h-title-wrap">
<span class="h2h-pill-success">OKF Memory Pattern</span>
<h4>Progressive Disclosure (In-Memory BM25)</h4>
</div>
<span class="h2h-score score-glow" id="h2h-okf-tokens">917 Tokens</span>
</div>
<p class="h2h-desc">
In-memory BM25 index (<300µs) extracts <em>only</em> the single verified ADR relevant to the user's specific coding task.
</p>
<ul class="h2h-list">
<li><span class="check">✓</span> <strong>Instant First Token:</strong> <span id="h2h-okf-ttft">11,565 ms</span> (<span id="h2h-okf-speedup">2.4x faster</span>)</li>
<li><span class="check">✓</span> <strong>O(1) Bounded Prompt:</strong> Consistently minimal context overhead</li>
<li><span class="check">✓</span> <strong>Zero Distraction:</strong> LLM focuses purely on the target policy decision</li>
<li><span class="check">✓</span> <strong>100% Policy Adherence:</strong> Flawless code generation with zero hallucinations</li>
</ul>
</div>
</div>
<!-- Reproduce Banner -->
<div class="benchmark-repro-card">
<div class="repro-left">
<div class="repro-title">Reproduce These Benchmarks On Your Own Machine</div>
<p class="repro-desc">100% pure Go standard library suite. Zero external dependencies. Supports LM Studio, Ollama, OpenAI, Claude, and Gemini.</p>
</div>
<div class="repro-cmd-box" data-copy="make benchmark">
<code>$ make benchmark</code>
<span class="copy-icon">📋</span>
</div>
</div>
<!-- Architectural Feature Matrix -->
<div class="matrix-header-wrap">
<h3 class="matrix-sub-title">System Architecture Matrix: OKF vs. Alternatives</h3>
<p class="matrix-sub-desc">Comparison of runtime latency, storage transparency, and operational complexity.</p>
</div>
<div class="table-wrapper">
<table class="benchmark-table">
<thead>
<tr>
<th>Feature / Metric</th>
<th class="col-highlight">OKF Agent Memory</th>
<th>Mem0 / Letta</th>
<th>Ad-hoc Markdown (CLAUDE.md)</th>
</tr>
</thead>
<tbody>
<tr>
<td><strong>Search Latency</strong></td>
<td class="col-highlight"><span class="badge-winner">< 300 µs</span> (In-Memory BM25)</td>
<td>150 – 800 ms (Vector API)</td>
<td>N/A (Full context scan)</td>
</tr>
<tr>
<td><strong>Cold Start / Overhead</strong></td>
<td class="col-highlight"><span class="badge-winner">< 4 ms</span> (Pure Go Binary)</td>
<td>1.5 – 3.0 s (Python / DB)</td>
<td>0 ms (Static File)</td>
</tr>
<tr>
<td><strong>Storage & Format</strong></td>
<td class="col-highlight">Standard OKF v0.2 Markdown</td>
<td>Postgres / Pinecone (Black Box)</td>
<td>Unstructured text</td>
</tr>
<tr>
<td><strong>Runtime Cost</strong></td>
<td class="col-highlight"><span class="badge-winner">0.00 €</span> (Zero API fees)</td>
<td>$$$ per Embedding & DB host</td>
<td>Context Bloat Costs</td>
</tr>
<tr>
<td><strong>Trust Signals</strong></td>
<td class="col-highlight"><code>verified:</code> vs <code>generated:</code></td>
<td>Opaque Vector Weights</td>
<td>None</td>
</tr>
<tr>
<td><strong>Git Native (Diff/Blame)</strong></td>
<td class="col-highlight">✅ 100% Native</td>
<td>❌ Proprietary DB</td>
<td>✅ Native</td>
</tr>
</tbody>
</table>
</div>
</div>
</section>
<!-- Architecture Visualizer -->
<section id="architecture" class="section">
<div class="container">
<div class="section-header">
<span class="section-tag">Design System</span>
<h2 class="section-title">The 5-Layer Architectural Model</h2>
<p class="section-desc">A clean, decoupled hierarchy ensuring portability across any AI agent or IDE.</p>
</div>
<div class="arch-stack">
<div class="arch-layer">
<span class="layer-num">LAYER 5</span>
<div class="layer-info">
<h4>Knowledge Corpus Layer (knowledge/)</h4>
<p>Conformant OKF v0.2 bundle containing dated change logs, indexes, and frontmatter concepts.</p>
</div>
</div>
<div class="arch-layer">
<span class="layer-num">LAYER 4</span>
<div class="layer-info">
<h4>Deterministic Tooling Layer (Go CLI & MCP)</h4>
<p>Zero-dependency single executable providing sub-300µs BM25 search, mutations, and graph validation.</p>
</div>
</div>
<div class="arch-layer">
<span class="layer-num">LAYER 3</span>
<div class="layer-info">
<h4>Agent Skills Layer (.agents/skills/okf-memory/)</h4>
<p>Structured operational playbooks for Claude Code, Cursor, Windsurf, and custom subagents.</p>
</div>
</div>
<div class="arch-layer">
<span class="layer-num">LAYER 2</span>
<div class="layer-info">
<h4>Convention Layer (Convention v0.1)</h4>
<p>Normative behavioral rules: Search-Before-Write, Anti-Duplication, and Trust Tiers.</p>
</div>
</div>
<div class="arch-layer">
<span class="layer-num">LAYER 1</span>
<div class="layer-info">
<h4>Specification Layer (Open Knowledge Format v0.2)</h4>
<p>Standardized YAML frontmatter and Markdown syntax defined by Google Cloud.</p>
</div>
</div>
</div>
</div>
</section>
<!-- Quickstart Section -->
<section id="quickstart" class="section">
<div class="container">
<div class="section-header">
<span class="section-tag">Zero to Productive</span>
<h2 class="section-title">Get Started in 60 Seconds</h2>
<p class="section-desc">Install via universal cURL 1-liner or Go and bootstrap any repository with a single command.</p>
</div>
<div class="features-grid">
<div class="feature-card">
<div class="feature-icon">1️⃣</div>
<h3 class="feature-title">Install OKF CLI</h3>
<p class="feature-text">Universal 1-liner for macOS (Apple Silicon / Intel) and Linux:</p>
<div class="install-pill" style="margin-top: 16px; width: 100%;"
data-copy="curl -fsSL https://okf-memory.dev/install.sh | sh">
<span>$ curl -fsSL https://okf-memory.dev/install.sh | sh</span>
<span class="copy-icon">📋</span>
</div>
</div>
<div class="feature-card">
<div class="feature-icon">2️⃣</div>
<h3 class="feature-title">Bootstrap Any Project</h3>
<p class="feature-text">Switch into your project and initialize memory:</p>
<div class="install-pill" style="margin-top: 16px; width: 100%;" data-copy="okf bootstrap .">
<span>$ okf bootstrap .</span>
<span class="copy-icon">📋</span>
</div>
</div>
<div class="feature-card">
<div class="feature-icon">3️⃣</div>
<h3 class="feature-title">Connect Your Agent</h3>
<p class="feature-text">Connect Claude Code, Cursor, or Windsurf directly via the built-in stdio MCP server.
</p>
</div>
</div>
<!-- Interactive MCP Config Viewer -->
<div class="mcp-config-box">
<div class="mcp-header">
<div class="mcp-tabs">
<button class="mcp-tab-btn active" data-mcp="claude">Claude Desktop / Code</button>
<button class="mcp-tab-btn" data-mcp="cursor">Cursor (.cursor/mcp.json)</button>
<button class="mcp-tab-btn" data-mcp="windsurf">Windsurf</button>
<button class="mcp-tab-btn" data-mcp="cli">Agent Direct (AGENTS.md)</button>
</div>
<button class="mcp-copy-btn" id="mcp-copy-trigger">📋 Copy Config</button>
</div>
<div class="mcp-body">
<pre class="mcp-code" id="mcp-code-display">{
"mcpServers": {
"okf-memory": {
"command": "okf",
"args": ["mcp", "knowledge"]
}
}
}</pre>
</div>
</div>
</div>
</section>
<!-- FAQ Section -->
<section id="faq" class="section">
<div class="container">
<div class="section-header">
<span class="section-tag">Common Questions</span>
<h2 class="section-title">Frequently Asked Questions</h2>
<p class="section-desc">Everything you need to know about OKF Agent Memory, architecture, and deployment.</p>
</div>
<div class="faq-accordion">
<div class="faq-item">
<button class="faq-trigger">
<span>Is OKF Agent Memory an official Google product?</span>
<span class="faq-icon">+</span>
</button>
<div class="faq-content">
<p>No. Google Cloud created and published the open <strong>Open Knowledge Format (OKF) v0.2</strong>
specification. <strong>OKF Agent Memory</strong> is an independent open-source Go implementation that
standardizes OKF as a persistent memory layer for AI agents with sub-300µs BM25 search, CLI tooling, and
an embedded Model Context Protocol (MCP) server.</p>
</div>
</div>
<div class="faq-item">
<button class="faq-trigger">
<span>Why in-memory BM25 search instead of a Vector Database (Mem0, Letta)?</span>
<span class="faq-icon">+</span>
</button>
<div class="faq-content">
<p>Vector databases require background containers, Python runtimes, and costly API embedding calls for
every write and query, adding 150ms–800ms of latency. OKF loads your repository's knowledge bundle
directly into memory in <4ms and searches with BM25 in <300µs — with 0.00 € API costs and 100%
offline predictability.</p>
</div>
</div>
<div class="faq-item">
<button class="faq-trigger">
<span>How does OKF prevent "Memory Rot" compared to CLAUDE.md or .cursorrules?</span>
<span class="faq-icon">+</span>
</button>
<div class="faq-content">
<p>Flat markdown files inevitably grow into 20,000-token monoliths that blow your context window and
degrade agent intelligence. OKF employs <strong>Progressive Disclosure</strong>: a structured graph of
atomic, interlinked concept files (<code>knowledge/</code>). The agent queries the index and loads only
the exact 300-token concept it needs. Plus, <code>okf validate --drift</code> continuously ensures links
and summaries stay synchronized.</p>
</div>
</div>
<div class="faq-item">
<button class="faq-trigger">
<span>How is this different from session loggers like agentmemory (agent-memory.dev)?</span>
<span class="faq-icon">+</span>
</button>
<div class="faq-content">
<p>Tools like <code>agentmemory</code> run a local Node.js daemon that intercepts chat transcripts and
attempts to summarize session logs with background LLM API calls. In contrast, <strong>OKF Agent
Memory</strong> follows the Karpathy "LLM Wiki" principle: curated, high-signal architecture decisions
authored directly into Git Markdown, version-controlled via Pull Requests, with strict Trust Tiers
(<code>verified: human</code> vs. <code>generated: agent</code>) and zero running daemon overhead.</p>
</div>
</div>
<div class="faq-item">
<button class="faq-trigger">
<span>Can I use OKF completely offline with local models (Ollama, LM Studio)?</span>
<span class="faq-icon">+</span>
</button>
<div class="faq-content">
<p>Yes, 100%. The Go single binary runs entirely on your local machine with zero external network calls
and zero telemetry. It is especially effective for local 8B–14B models (e.g. Qwen 2.5 Coder, Llama 3)
because it drastically reduces context prefill latency from 6+ seconds down to under 50ms.</p>
</div>
</div>
<div class="faq-item">
<button class="faq-trigger">
<span>How does it handle multi-repository enterprise setups?</span>
<span class="faq-icon">+</span>
</button>
<div class="faq-content">
<p>In open-source, each repository manages its own self-contained memory bundle in
<code>knowledge/</code>. For engineering teams with 50+ microservices, our upcoming <strong>OKF
Cloud</strong> enables federated cross-repo search, visual graph exploration, and automated PR
governance bots.
</p>
</div>
</div>
</div>
</div>
</section>
<!-- Waitlist / Dual Launch Call to Action -->
<section id="waitlist" class="section">
<div class="container">
<div class="waitlist-card">
<span class="section-tag" style="color: #c4b5fd;">Enterprise & Teams</span>
<h2 class="section-title">Need Cross-Repo Memory & Visual Graphs?</h2>
<p class="section-desc" style="max-width: 600px; margin: 0 auto;">
Join the <strong>OKF Cloud Private Beta</strong>. Connect 50+ microservices into a unified federated memory
graph, with human-in-the-loop verification portals and automated PR compliance bots.
</p>
<form id="waitlist-form" class="waitlist-form" data-provider="brevo"
data-brevo-action="https://861acd5a.sibforms.com/serve/MUIFAKnhOMBlsXCV2K6cq0zhN39PEsWAj0gQrEjvnyOJuGDWus7p2t45V7hFEYEaaRXP-xPCU3lodz3ACE3OrUhW6ZpG-icA5MS-xZuSNFUgOk7GBjykxVnzgwDx9tKCHw22OYJxSbApIRp-NoqGKjoR97LnCFjHW_QgZalZrgjGXsnPppbeeT4gSblO14Xkoc7-lWKrt_WnFnCMPQ==">
<input type="email" id="waitlist-email" class="waitlist-input"
placeholder="Enter your work email (e.g. alex@company.com)" required>
<button type="submit" class="btn btn-primary" style="padding: 14px 28px;">Request VIP Access</button>
<div class="form-consent">
<input type="checkbox" id="waitlist-consent" required>
<label for="waitlist-consent">I agree to receive the OKF Cloud Beta invitation and early developer
updates. (Unsubscribe anytime. Powered by Brevo / EU data-protection compliant).</label>
</div>
</form>
<div id="waitlist-feedback" style="display: none;"></div>
<p style="font-size: 0.8rem; color: var(--text-muted); margin-top: 12px;">No spam. Early beta invitations &
developer preview updates only.</p>
</div>
</div>
</section>
</main>
<!-- Footer -->
<footer class="site-footer">
<div class="container footer-wrap">
<div>
<p><strong>OKF Agent Memory</strong> — Open Knowledge Format (OKF) v0.2 Persistent Project Memory.</p>
<p style="margin-top: 4px;">Software Release <a
href="https://github.com/okf-memory/okf-agent-memory/releases/tag/v0.6.1" target="_blank" rel="noopener"
style="color: #38bdf8; text-decoration: underline;">v0.6.1</a> under the permissive MIT License.</p>
</div>
<ul class="footer-links">
<li><a href="https://github.com/okf-memory/okf-agent-memory" target="_blank" rel="noopener">GitHub</a></li>
<li><a href="https://registry.okf-memory.dev" target="_blank" rel="noopener">Bundle Registry</a></li>
<li><a href="https://github.com/okf-memory/okf-agent-memory/releases/tag/v0.6.1" target="_blank"
rel="noopener">Release v0.6.1</a></li>
<li><a href="https://github.com/okf-memory/okf-agent-memory/blob/main/docs/spec/CONVENTION.md" target="_blank"
rel="noopener">Convention v0.1</a></li>
<li><a href="https://github.com/okf-memory/okf-agent-memory/blob/main/docs/guides/CLI.md" target="_blank"
rel="noopener">CLI Docs</a></li>
<li><a href="https://github.com/okf-memory/homebrew-tap" target="_blank" rel="noopener">Homebrew Tap</a></li>
<li><a href="https://github.com/okf-memory/okf-agent-memory/blob/main/docs/guides/GETTING_STARTED.md" target="_blank"
rel="noopener">Install Guide</a></li>
<li><button id="cookie-settings-btn" class="footer-link-btn" type="button">Cookie Settings</button></li>
</ul>
</div>
</footer>
<!-- Cookie Consent Banner (DSGVO / GDPR Compliant) -->
<div id="cookie-banner" class="cookie-banner" role="dialog" aria-modal="true" aria-label="Privacy & Cookie Preferences" style="display: none;">
<div class="cookie-banner-inner">
<div class="cookie-banner-text">
<div class="cookie-banner-header">
<span class="cookie-icon">🍪</span>
<strong>Privacy & Cookie Preferences</strong>
</div>
<p>
We use privacy-friendly Google Analytics 4 (anonymized IP) to understand how developers discover OKF Agent Memory. No advertising cookies are used. You can adjust or revoke your choice anytime.
</p>
</div>
<div class="cookie-banner-actions">
<button id="cookie-btn-reject" class="btn btn-cookie-secondary" type="button">Essential Only</button>
<button id="cookie-btn-accept" class="btn btn-cookie-primary" type="button">Accept Analytics</button>
</div>
</div>
</div>
<script src="app.js"></script>
</body>
</html>