feat: auto-track all Claude Code sessions via ANTHROPIC_BASE_URL proxy

- Add POST /v1/messages (Anthropic-format passthrough) to core with full
  ticket tracking — every Claude Code request creates a TokenVault ticket
- Add GET /v1/models passthrough so Anthropic SDK model listing works
- Proxy /v1/messages + /v1/models in dashboard so public URL forwards to core
- Add /health endpoint to dashboard for infra monitoring
- Overview tab now shows RTK savings banner (3.5M tokens, 753 commands)
  alongside LLM ticket stats — Overview no longer looks empty
- Provider split bar chart rendered when request data is available

ANTHROPIC_BASE_URL=https://tokenvault.fichtmueller.org now set globally
in ~/.zshrc — every new Claude Code session is tracked automatically.

packages/core/src/routes/anthropic-proxy.ts

@@ -0,0 +1,159 @@
+import type { FastifyInstance } from 'fastify';
+import { createTicket, hashContent } from '../tickets/ticket-service.js';
+import { logger } from '../observability/logger.js';
+import { config } from '../config.js';
+
+// Anthropic pricing per 1M tokens (USD)
+const ANTHROPIC_PRICING: Record<string, { input: number; output: number; cached?: number }> = {
+  'claude-opus-4-20250514':    { input: 15.0,  output: 75.0,  cached: 1.50  },
+  'claude-opus-4-5':           { input: 15.0,  output: 75.0,  cached: 1.50  },
+  'claude-sonnet-4-20250514':  { input: 3.0,   output: 15.0,  cached: 0.30  },
+  'claude-sonnet-4-5-20251001':{ input: 3.0,   output: 15.0,  cached: 0.30  },
+  'claude-haiku-3-5-20251022': { input: 0.80,  output: 4.0,   cached: 0.08  },
+  'claude-haiku-3-20250307':   { input: 0.25,  output: 1.25,  cached: 0.025 },
+};
+
+function calcCost(model: string, inputTokens: number, outputTokens: number, cachedTokens: number): number {
+  const pricing = ANTHROPIC_PRICING[model] ?? ANTHROPIC_PRICING['claude-sonnet-4-20250514']!;
+  const inputCost  = ((inputTokens - cachedTokens) / 1_000_000) * pricing.input;
+  const outputCost = (outputTokens / 1_000_000) * pricing.output;
+  const cacheCost  = pricing.cached ? (cachedTokens / 1_000_000) * pricing.cached : 0;
+  return Math.max(0, inputCost + outputCost + cacheCost);
+}
+
+/**
+ * Anthropic API passthrough — enables ANTHROPIC_BASE_URL routing.
+ *
+ * Claude Code (and any Anthropic SDK client) sends requests here when
+ * ANTHROPIC_BASE_URL=https://tokenvault.fichtmueller.org is set.
+ * We forward the request to Anthropic using the client's own API key,
+ * then create a TokenVault ticket to track usage and cost.
+ */
+export async function anthropicProxyRoutes(app: FastifyInstance): Promise<void> {
+
+  // ─── POST /v1/messages — Anthropic Messages API passthrough ──────────────
+  app.post('/v1/messages', async (req, reply) => {
+    const body = req.body as Record<string, unknown>;
+    const model = (body['model'] as string) ?? 'claude-sonnet-4-20250514';
+
+    // Use the client's API key if provided, fall back to configured key
+    const apiKey = (req.headers['x-api-key'] as string)
+      || (req.headers['authorization'] as string)?.replace(/^Bearer /, '')
+      || config.providers.anthropic.apiKey;
+
+    const anthropicVersion = (req.headers['anthropic-version'] as string) ?? '2023-06-01';
+    const anthropicBeta    = req.headers['anthropic-beta'] as string | undefined;
+
+    const caller  = req.headers['x-tokenvault-caller'] as string | undefined;
+    const project = req.headers['x-tokenvault-project'] as string | undefined;
+    const team    = req.headers['x-tokenvault-team'] as string | undefined;
+
+    const start = Date.now();
+
+    // ── Forward to Anthropic ────────────────────────────────────────────────
+    const forwardHeaders: Record<string, string> = {
+      'Content-Type':       'application/json',
+      'x-api-key':          apiKey,
+      'anthropic-version':  anthropicVersion,
+    };
+    if (anthropicBeta) forwardHeaders['anthropic-beta'] = anthropicBeta;
+
+    const upstream = await fetch('https://api.anthropic.com/v1/messages', {
+      method:  'POST',
+      headers: forwardHeaders,
+      body:    JSON.stringify(body),
+    });
+
+    const latency = Date.now() - start;
+    const responseBody = await upstream.json() as Record<string, unknown>;
+
+    // ── Track as TokenVault ticket (best-effort, non-blocking) ─────────────
+    if (upstream.ok) {
+      const usage = responseBody['usage'] as {
+        input_tokens?: number;
+        output_tokens?: number;
+        cache_read_input_tokens?: number;
+        cache_creation_input_tokens?: number;
+      } | undefined;
+
+      if (usage) {
+        const inputTokens  = usage.input_tokens  ?? 0;
+        const outputTokens = usage.output_tokens ?? 0;
+        const cachedTokens = usage.cache_read_input_tokens ?? 0;
+        const cost = calcCost(model, inputTokens, outputTokens, cachedTokens);
+
+        const messages = (body['messages'] as Array<{ role: string; content: unknown }>) ?? [];
+        const inputText = messages.map(m => {
+          if (typeof m.content === 'string') return m.content;
+          if (Array.isArray(m.content)) return m.content.map((c: Record<string, unknown>) => c['text'] ?? '').join('');
+          return '';
+        }).join('\n');
+        const outputContent = (responseBody['content'] as Array<{ text?: string }> | undefined) ?? [];
+        const outputText = outputContent.map(c => c.text ?? '').join('');
+
+        createTicket({
+          provider:   'anthropic',
+          model,
+          status:     'completed',
+          tokens_in:  inputTokens,
+          tokens_out: outputTokens,
+          tokens_cached: cachedTokens,
+          tokens_saved:  0,
+          cost_usd:   cost,
+          latency_ms: latency,
+          cache_hit:  cachedTokens > 0,
+          caller:     caller ?? 'claude-code',
+          project,
+          team,
+          input_hash:  hashContent(inputText),
+          output_hash: hashContent(outputText),
+        }).catch(err => logger.warn({ err }, 'Failed to create ticket for Anthropic passthrough'));
+      }
+    } else {
+      // Track failed requests too
+      createTicket({
+        provider:   'anthropic',
+        model,
+        status:     'failed',
+        tokens_in:  0,
+        tokens_out: 0,
+        cost_usd:   0,
+        latency_ms: latency,
+        caller:     caller ?? 'claude-code',
+        project,
+        team,
+        input_hash: hashContent(model),
+        output_hash: '',
+      }).catch(() => { /* best effort */ });
+
+      logger.warn({ model, status: upstream.status }, 'Anthropic upstream error');
+    }
+
+    // ── Return Anthropic response as-is ────────────────────────────────────
+    reply.code(upstream.status);
+    reply.header('Content-Type', 'application/json');
+
+    // Pass through useful Anthropic response headers
+    const anthropicReqId = upstream.headers.get('request-id');
+    if (anthropicReqId) reply.header('request-id', anthropicReqId);
+
+    return responseBody;
+  });
+
+  // ─── GET /v1/models — Anthropic models list passthrough ─────────────────
+  app.get('/v1/models', async (req, reply) => {
+    const apiKey = (req.headers['x-api-key'] as string)
+      || config.providers.anthropic.apiKey;
+
+    const upstream = await fetch('https://api.anthropic.com/v1/models', {
+      headers: {
+        'x-api-key':         apiKey,
+        'anthropic-version': '2023-06-01',
+      },
+    });
+
+    reply.code(upstream.status);
+    reply.header('Content-Type', 'application/json');
+    return upstream.json();
+  });
+}

packages/core/src/server.ts

@@ -8,6 +8,7 @@ import { closePool } from './db/client.js';
 import { initProviders } from './providers/index.js';
 import { healthRoutes } from './routes/health.js';
 import { proxyRoutes } from './routes/proxy.js';
+import { anthropicProxyRoutes } from './routes/anthropic-proxy.js';
 import { ticketRoutes } from './routes/tickets.js';
 import { rtkRoutes } from './routes/rtk.js';
 
@@ -20,6 +21,7 @@ await app.register(cors, { origin: true });
 // ─── Routes ──────────────────────────────────────────────────────────────────
 await app.register(healthRoutes);
 await app.register(proxyRoutes);
+await app.register(anthropicProxyRoutes);
 await app.register(ticketRoutes);
 await app.register(rtkRoutes);
 

packages/dashboard/public/index.html

@@ -161,9 +161,21 @@ tr:hover { background:#f8fafc; }
     <div class="cards" id="stats-cards">
       <div class="loading"><div class="spin"></div> Loading...</div>
     </div>
+    <!-- RTK savings banner — always visible since RTK data exists independently of LLM routing -->
+    <div id="rtk-overview-banner" style="display:none;margin:16px 0;background:linear-gradient(135deg,#f0fdf4,#dcfce7);border:1px solid #86efac;border-radius:12px;padding:20px 24px">
+      <div style="display:flex;align-items:center;gap:12px;margin-bottom:12px">
+        <span style="font-size:20px">⚡</span>
+        <strong style="color:#16a34a;font-size:15px">RTK Token Savings (all-time)</strong>
+        <span style="margin-left:auto;font-size:12px;color:#15803d" id="rtk-banner-host"></span>
+      </div>
+      <div style="display:grid;grid-template-columns:repeat(auto-fit,minmax(150px,1fr));gap:12px" id="rtk-banner-cards"></div>
+    </div>
     <div class="chart-row">
       <div class="chart-placeholder" id="cost-timeline">Cost timeline will appear here after requests are tracked</div>
-      <div class="chart-placeholder" id="provider-split">Provider split will appear here</div>
+      <div id="provider-split" style="background:var(--surface);border:1px solid var(--border);border-radius:12px;padding:24px;flex:1;min-height:180px">
+        <div style="font-weight:600;margin-bottom:12px;color:var(--text-muted)">Provider Split</div>
+        <div id="provider-split-inner" style="color:var(--text-muted);font-size:13px;text-align:center;margin-top:40px">Provider split will appear here after requests are tracked</div>
+      </div>
     </div>
   </div>
 
@@ -311,7 +323,10 @@ function fmtTime(d) { return new Date(d).toLocaleString(lang === 'de' ? 'de-DE'
 
 async function loadStats() {
   try {
-    const data = await (await fetch(API + '/tickets/stats?period=month')).json();
+    const [data, rtkStats] = await Promise.all([
+      (await fetch(API + '/tickets/stats?period=month')).json(),
+      fetch(API + '/rtk/stats?period=all').then(r => r.ok ? r.json() : null).catch(() => null),
+    ]);
     const labels = t[lang];
     document.getElementById('stats-cards').innerHTML = `
       <div class="card"><div class="label">${labels.totalCost}</div><div class="value primary">$${fmt(data.total_cost_usd)}</div><div class="sub">Last 30 days</div></div>
@@ -321,6 +336,53 @@ async function loadStats() {
       <div class="card"><div class="label">${labels.tokensIn}</div><div class="value">${fmtK(data.total_tokens_in)}</div></div>
       <div class="card"><div class="label">${labels.tokensOut}</div><div class="value">${fmtK(data.total_tokens_out)}</div></div>
     `;
+
+    // ── RTK savings banner ──────────────────────────────────────────────────
+    if (rtkStats && rtkStats.total_commands > 0) {
+      const banner = document.getElementById('rtk-overview-banner');
+      banner.style.display = 'block';
+      document.getElementById('rtk-banner-host').textContent =
+        `${rtkStats.unique_hosts} host${rtkStats.unique_hosts !== 1 ? 's' : ''} · ${rtkStats.unique_projects} project${rtkStats.unique_projects !== 1 ? 's' : ''}`;
+      document.getElementById('rtk-banner-cards').innerHTML = `
+        <div style="background:#fff;border-radius:8px;padding:12px;text-align:center">
+          <div style="font-size:11px;color:#6b7280;margin-bottom:4px">TOKENS SAVED</div>
+          <div style="font-size:22px;font-weight:700;color:#16a34a">${fmtK(rtkStats.total_saved_tokens)}</div>
+        </div>
+        <div style="background:#fff;border-radius:8px;padding:12px;text-align:center">
+          <div style="font-size:11px;color:#6b7280;margin-bottom:4px">COMMANDS</div>
+          <div style="font-size:22px;font-weight:700;color:#6366f1">${fmtK(rtkStats.total_commands)}</div>
+        </div>
+        <div style="background:#fff;border-radius:8px;padding:12px;text-align:center">
+          <div style="font-size:11px;color:#6b7280;margin-bottom:4px">AVG SAVINGS</div>
+          <div style="font-size:22px;font-weight:700;color:#f59e0b">${rtkStats.avg_savings_pct.toFixed(1)}%</div>
+        </div>
+        <div style="background:#fff;border-radius:8px;padding:12px;text-align:center">
+          <div style="font-size:11px;color:#6b7280;margin-bottom:4px">INPUT TOKENS</div>
+          <div style="font-size:22px;font-weight:700">${fmtK(rtkStats.total_input_tokens)}</div>
+        </div>
+      `;
+    }
+
+    // ── Provider split (simple bar chart) ───────────────────────────────────
+    try {
+      const breakdown = await (await fetch(API + '/cost/breakdown?group_by=provider')).json();
+      const total = breakdown.reduce((s, b) => s + b.request_count, 0);
+      if (total > 0) {
+        const colors = { anthropic:'#6366f1', openai:'#10b981', ollama:'#f59e0b', 'ai-bridge':'#3b82f6' };
+        document.getElementById('provider-split-inner').innerHTML = breakdown.map(b => {
+          const pct = ((b.request_count / total) * 100).toFixed(0);
+          const col = colors[b.group_value] || '#94a3b8';
+          return `<div style="margin-bottom:10px">
+            <div style="display:flex;justify-content:space-between;font-size:12px;margin-bottom:3px">
+              <span style="font-weight:500">${b.group_value}</span><span style="color:var(--text-muted)">${b.request_count} req · ${pct}%</span>
+            </div>
+            <div style="background:var(--border);border-radius:4px;height:6px">
+              <div style="background:${col};width:${pct}%;height:6px;border-radius:4px"></div>
+            </div>
+          </div>`;
+        }).join('');
+      }
+    } catch {}
   } catch { document.getElementById('stats-cards').innerHTML = '<div class="card"><div class="label">Status</div><div class="value">Offline</div><div class="sub">TokenVault core not reachable</div></div>'; }
 }
 

packages/dashboard/src/server.ts

@@ -63,5 +63,42 @@ app.get('/api/rtk/hosts', async () => {
   return res.json();
 });
 
+// ─── Anthropic API passthrough (ANTHROPIC_BASE_URL routing) ──────────────────
+// Claude Code routes through here when ANTHROPIC_BASE_URL=https://tokenvault.fichtmueller.org
+// Forward to core which handles tracking + proxying to Anthropic.
+app.post('/v1/messages', async (req, reply) => {
+  const headers: Record<string, string> = { 'Content-Type': 'application/json' };
+  for (const key of ['x-api-key', 'anthropic-version', 'anthropic-beta', 'x-tokenvault-caller', 'x-tokenvault-project', 'x-tokenvault-team']) {
+    const val = req.headers[key];
+    if (val) headers[key] = val as string;
+  }
+  const res = await fetch(`${CORE_URL}/v1/messages`, {
+    method: 'POST',
+    headers,
+    body: JSON.stringify(req.body),
+  });
+  reply.code(res.status);
+  reply.header('Content-Type', 'application/json');
+  const reqId = res.headers.get('request-id');
+  if (reqId) reply.header('request-id', reqId);
+  return res.json();
+});
+
+app.get('/v1/models', async (req, reply) => {
+  const headers: Record<string, string> = {};
+  const apiKey = req.headers['x-api-key'];
+  if (apiKey) headers['x-api-key'] = apiKey as string;
+  const res = await fetch(`${CORE_URL}/v1/models`, { headers });
+  reply.code(res.status);
+  reply.header('Content-Type', 'application/json');
+  return res.json();
+});
+
+// ─── Health endpoint (needed for infra monitoring) ────────────────────────────
+app.get('/health', async () => {
+  const res = await fetch(`${CORE_URL}/health`);
+  return res.json();
+});
+
 await app.listen({ port: PORT, host: '0.0.0.0' });
 console.log(`TokenVault Dashboard running on http://localhost:${PORT}`);