File size: 9,648 Bytes
8f16a6b | 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 | <!DOCTYPE html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>AgentBridge β Self-hosted AI office agents</title>
<meta name="description" content="AgentBridge: self-hosted AI agents for office work β terminal chat, OpenAI-compatible API and MCP in one process. Documents, spreadsheets, email, web research, voice, SIP phone and Telegram.">
<style>
:root { --blue:#1f6feb; --green:#2ea043; --bg:#0b1120; --panel:#111a2e; --text:#dbe4f3; --muted:#8aa0c4; --line:#233458; }
* { box-sizing:border-box; margin:0; padding:0; }
body { background:var(--bg); color:var(--text); font:16px/1.6 -apple-system,"Segoe UI",Roboto,Helvetica,Arial,sans-serif; }
.wrap { max-width:1080px; margin:0 auto; padding:0 24px; }
header { border-bottom:1px solid var(--line); background:linear-gradient(180deg,#0d1530,var(--bg)); }
nav { display:flex; align-items:center; gap:20px; padding:14px 0; }
nav .logo { font-weight:700; font-size:18px; color:#fff; }
nav a { color:var(--muted); text-decoration:none; font-size:14px; }
nav a:hover { color:#fff; }
nav .spacer { flex:1; }
.hero { padding:64px 0 56px; text-align:center; }
.hero h1 { font-size:44px; line-height:1.15; color:#fff; }
.hero h1 span { background:linear-gradient(90deg,var(--blue),var(--green)); -webkit-background-clip:text; background-clip:text; color:transparent; }
.hero p.tag { max-width:760px; margin:18px auto 30px; color:var(--muted); font-size:18px; }
.btn { display:inline-block; padding:12px 26px; border-radius:8px; font-weight:600; text-decoration:none; margin:0 6px; }
.btn.primary { background:var(--blue); color:#fff; }
.btn.ghost { border:1px solid var(--line); color:var(--text); }
.btn:hover { filter:brightness(1.12); }
.hero img.poster { width:100%; max-width:760px; border-radius:12px; margin-top:40px; border:1px solid var(--line); }
section { padding:56px 0; border-top:1px solid var(--line); }
h2 { font-size:28px; color:#fff; margin-bottom:24px; }
.grid { display:grid; grid-template-columns:repeat(auto-fit,minmax(300px,1fr)); gap:16px; }
.card { background:var(--panel); border:1px solid var(--line); border-radius:12px; padding:20px; }
.card h3 { font-size:17px; color:#fff; margin-bottom:8px; }
.card p { color:var(--muted); font-size:14.5px; }
ul.docs { columns:2; column-gap:40px; list-style:none; }
ul.docs li { padding:6px 0; border-bottom:1px dashed var(--line); break-inside:avoid; }
ul.docs a { color:var(--text); text-decoration:none; }
ul.docs a:hover { color:var(--blue); }
.demo { display:grid; grid-template-columns:repeat(auto-fit,minmax(280px,1fr)); gap:16px; }
.demo figure { background:var(--panel); border:1px solid var(--line); border-radius:12px; overflow:hidden; }
.demo img, .demo video { width:100%; display:block; aspect-ratio:16/9; object-fit:cover; background:#000; }
.demo figcaption { padding:10px 14px; font-size:13.5px; color:var(--muted); }
.wp { background:linear-gradient(135deg,#101c3a,#0f2a1e); border:1px solid var(--line); border-radius:12px; padding:26px; display:flex; gap:22px; align-items:center; flex-wrap:wrap; }
.wp .txt { flex:1; min-width:280px; }
.wp h3 { color:#fff; font-size:20px; margin-bottom:6px; }
.wp p { color:var(--muted); }
.wp a { color:var(--green); }
footer { border-top:1px solid var(--line); padding:28px 0 40px; text-align:center; color:var(--muted); font-size:13.5px; }
footer a { color:var(--muted); }
@media (max-width:640px){ .hero h1{font-size:32px} ul.docs{columns:1} }
</style>
</head>
<body>
<header>
<div class="wrap">
<nav>
<span class="logo">π€ AgentBridge</span>
<span class="spacer"></span>
<a href="#features">Features</a>
<a href="#docs">Docs</a>
<a href="#white-paper">White paper</a>
<a href="#demos">Demos</a>
<a href="https://github.com/Graphene-Lab/AgentBridge">GitHub β</a>
</nav>
</div>
</header>
<div class="hero">
<div class="wrap">
<h1>Self-hosted AI agents that <span>do real office work</span></h1>
<p class="tag">One process: a full-screen chat terminal, an OpenAI-compatible HTTP API and a native
MCP connector. Documents, spreadsheets, email, web research, voice, podcasts β while your data
never leaves your machine.</p>
<a class="btn primary" href="https://graphenelab.it/AgentBridge/download/">Download for your OS</a>
<a class="btn ghost" href="docs/MANUAL.md">Read the manual</a>
<img class="poster" src="media/agent-bridge-poster.png" alt="AgentBridge">
</div>
</div>
<section id="features">
<div class="wrap">
<h2>What it can do</h2>
<div class="grid">
<div class="card"><h3>π Real documents</h3><p>Word, Excel, PowerPoint and detailed PDF reports generated end-to-end from a plain prompt, from official templates.</p></div>
<div class="card"><h3>π§ Deterministic Memory</h3><p>The agent permanently knows the things, facts, people and documents it has handled β zero reasoning iterations wasted on recollection.</p></div>
<div class="card"><h3>π‘οΈ True sandbox</h3><p>An application-level action perimeter no tool method can breach. Structurally enforced by code, not by prompt.</p></div>
<div class="card"><h3>ποΈ Voice & phone</h3><p>Kokoro neural TTS out of the box, and a SIP phone endpoint reachable 24/7 behind a DTMF PIN.</p></div>
<div class="card"><h3>π¬ Telegram</h3><p>The agents live in your Telegram as a chat contact, attachments both ways, allow-list access control.</p></div>
<div class="card"><h3>ποΈ Autonomous scheduling</h3><p>Deadlines and recurring duties the agent plans and executes on its own β reports, checks, follow-ups.</p></div>
<div class="card"><h3>π MCP + OpenAI API</h3><p>Point any OpenAI client at <code>localhost:5290/v1</code>, or drive the same runtime through standard MCP clients.</p></div>
<div class="card"><h3>π GDPR-ready</h3><p>Optional anonymization masks personal data before it reaches any external LLM provider, and restores it in the reply.</p></div>
</div>
</div>
</section>
<section id="docs">
<div class="wrap">
<h2>Documentation</h2>
<ul class="docs">
<li><a href="docs/INDEX.md">π Beginner guides (15)</a></li>
<li><a href="docs/MANUAL.md">π User manual β start here</a></li>
<li><a href="docs/TUI.md">π₯οΈ Terminal UI reference</a></li>
<li><a href="docs/API.md">π HTTP API reference</a></li>
<li><a href="docs/telegram.md">π¬ Telegram chat</a></li>
<li><a href="docs/sip.md">π SIP telephony</a></li>
<li><a href="docs/TTS.md">π Text-to-speech</a></li>
<li><a href="docs/autoupdate.md">π Auto-update</a></li>
</ul>
</div>
</section>
<section id="white-paper">
<div class="wrap">
<h2>White paper</h2>
<div class="wp">
<div class="txt">
<h3>Why the AIOrchestrator library model outperforms traditional MCP deployments</h3>
<p>One runtime, one agent core, one consistent operational flow across terminal and API β
faster execution, lower integration overhead and stronger control than multi-server MCP setups.</p>
<p><a href="docs/AIORCHESTRATOR-WHITEPAPER.md">Read the white paper β</a></p>
</div>
<div class="txt">
<h3>Universal Tool System (UTS)</h3>
<p>Tools that run natively as compiled .NET assemblies β no interpreters, no HTTP, in-process
with zero overhead, hot-add plugins, one AnyCPU binary across Windows, Linux, macOS and iOS.</p>
<p><a href="https://github.com/Graphene-Lab/AgentHarness/blob/master/docs/universal-tool-system.md">Read the UTS guide β</a></p>
</div>
</div>
</div>
</section>
<section id="demos">
<div class="wrap">
<h2>Demo gallery</h2>
<div class="demo">
<figure><video controls src="media/demo.mp4"></video><figcaption>Terminal UI β streaming agent replies, command palette, status bar.</figcaption></figure>
<figure><img src="media/demo1.gif" alt="Presentation demo"><figcaption>PowerPoint-style presentations generated by the agent.</figcaption></figure>
<figure><video controls src="media/enanced_doc_demo.mp4"></video><figcaption>Detailed PDF market & financial analysis reports.</figcaption></figure>
<figure><img src="media/invoice.png" alt="Invoice"><figcaption>Invoice generated end-to-end (material check β DOCX).</figcaption></figure>
<figure><img src="media/contract.png" alt="Employment contract"><figcaption>Employment contract in Word, generated from a request.</figcaption></figure>
<figure><img src="media/spreadsheet-demo.png" alt="Spreadsheet"><figcaption>Excel workbook with data table + chart, single A4 page.</figcaption></figure>
<figure><video controls src="media/podcast-example.mp4"></video><figcaption>Podcast sample episode (MP4, ~42 s) β MP3 + RSS also produced.</figcaption></figure>
<figure><img src="media/telegram-agent.png" alt="Telegram chat"><figcaption>Telegram agent chat.</figcaption></figure>
</div>
</div>
</section>
<footer>
<div class="wrap">
<p><a href="https://github.com/Graphene-Lab/AgentBridge">Source code on GitHub</a> Β·
<a href="https://graphenelab.it/AgentBridge/">graphenelab.it/AgentBridge</a> Β·
<a href="https://github.com/Graphene-Lab/AgentBridge/blob/master/LICENSE.md">Andrea Bruno License 1.4</a></p>
<p style="margin-top:8px">Self-contained archives (~460 MB, no .NET needed) Β· Windows x64 Β· Linux x64/ARM64 Β· macOS Intel/Apple Silicon</p>
</div>
</footer>
</body>
</html>
|