feat: v2.5.12 — plan and quota for every AI tool, Grok Build and Kimi Code, Quit AgentHub (#65)

Usage panel (status bar):
- every account shows its plan (Claude subscriptionType, Codex
  chatgpt_plan_type, Cursor membershipType, Grok subscriptionTier)
- new quota rows: Cursor CLI (cursor.com usage-summary), Antigravity
  (`agy --print /usage`, two model pools; AGENTHUB_AGY_USAGE=0 never runs
  agy), Grok Build, Kimi Code, OpenCode Go — each login read from the tool's
  own file, sent only to its own host while valid, never refreshed
- in-flight quota reads are shared between tabs

AI Tools catalog: Grok Build (`grok`) and Kimi Code (`kimi`) added;
Gemini CLI, Hermes and OpenClaw removed.

Also: Quit AgentHub (#65, `agenthub stop`), status-area icon, Add to Chat
on an editor selection, Settings in three groups, JetBrains-style Changes
panel (ticked files/hunks, ignore, blame, file history), terminal font.

--no-verify: :3000 is taken by another dev server (lane-detect tests);
both suites run by hand — server 2636/2639 (the 3 are :3000), client
1131/1131.

server/db.js

@@ -239,6 +239,14 @@ db.exec(`
     baseline_web_search INTEGER NOT NULL DEFAULT 0,
     baseline_web_fetch INTEGER NOT NULL DEFAULT 0,
     baseline_code_execution INTEGER NOT NULL DEFAULT 0,
+    -- Subsets of the token columns spent by requests whose prompt was above
+    -- LONG_CONTEXT_TOKENS (server/lib/pricing-constants.js), priced at the
+    -- rule's long_* rates. Kept at their high-water mark like the hourly rows.
+    long_input_tokens INTEGER NOT NULL DEFAULT 0,
+    long_output_tokens INTEGER NOT NULL DEFAULT 0,
+    long_cache_read_tokens INTEGER NOT NULL DEFAULT 0,
+    long_cache_write_tokens INTEGER NOT NULL DEFAULT 0,
+    long_cache_write_1h_tokens INTEGER NOT NULL DEFAULT 0,
     PRIMARY KEY (session_id, model, speed, inference_geo, service_tier),
     FOREIGN KEY (session_id) REFERENCES sessions(id) ON DELETE CASCADE
   );
@@ -262,6 +270,14 @@ db.exec(`
     web_search_requests INTEGER NOT NULL DEFAULT 0,
     web_fetch_requests INTEGER NOT NULL DEFAULT 0,
     code_execution_requests INTEGER NOT NULL DEFAULT 0,
+    -- Subsets of the token columns spent by requests whose prompt was above
+    -- LONG_CONTEXT_TOKENS (server/lib/pricing-constants.js), priced at the
+    -- rule's long_* rates. Kept at their high-water mark like the hourly rows.
+    long_input_tokens INTEGER NOT NULL DEFAULT 0,
+    long_output_tokens INTEGER NOT NULL DEFAULT 0,
+    long_cache_read_tokens INTEGER NOT NULL DEFAULT 0,
+    long_cache_write_tokens INTEGER NOT NULL DEFAULT 0,
+    long_cache_write_1h_tokens INTEGER NOT NULL DEFAULT 0,
     PRIMARY KEY (session_id, hour, model, speed, inference_geo, service_tier),
     FOREIGN KEY (session_id) REFERENCES sessions(id) ON DELETE CASCADE
   );
@@ -279,6 +295,11 @@ db.exec(`
     -- caching multipliers (see server/lib/pricing-constants.js).
     fast_input_per_mtok REAL NOT NULL DEFAULT 0,
     fast_output_per_mtok REAL NOT NULL DEFAULT 0,
+    -- Rates for a request whose prompt is above LONG_CONTEXT_TOKENS (Haiku 5.5);
+    -- 0 = no long-context pricing. Cache rates are derived from long_input by
+    -- the same ratio, as fast mode's are.
+    long_input_per_mtok REAL NOT NULL DEFAULT 0,
+    long_output_per_mtok REAL NOT NULL DEFAULT 0,
     -- Time-limited introductory rates. When intro_until is set, usage on/before
     -- that date (YYYY-MM-DD) is priced at the intro_* rates and usage after it at
     -- the standard rates — so promo pricing (e.g. Claude Sonnet 5's launch
@@ -999,6 +1020,19 @@ try {
   setFast.run(30, 150, "claude-opus-4-6%");
 }
 
+// Migrate: long-context rate columns (0 = none; filled by the pricing sync or
+// DEFAULT_PRICING's top-up below).
+try {
+  db.prepare("SELECT long_input_per_mtok FROM model_pricing LIMIT 1").get();
+} catch {
+  db.prepare(
+    "ALTER TABLE model_pricing ADD COLUMN long_input_per_mtok REAL NOT NULL DEFAULT 0"
+  ).run();
+  db.prepare(
+    "ALTER TABLE model_pricing ADD COLUMN long_output_per_mtok REAL NOT NULL DEFAULT 0"
+  ).run();
+}
+
 // Migrate: add time-limited introductory-rate columns to model_pricing.
 // Usage on/before intro_until prices at the intro_* rates; usage after prices at
 // standard — so promo pricing (e.g. Claude Sonnet 5's launch discount) stays
@@ -1602,6 +1636,27 @@ try {
   db.pragma("foreign_keys = ON");
 }
 
+// Migrate: long-context subsets (Haiku 5.5 bills a prompt above 100k tokens at
+// higher rates). Plain added columns, NOT a new key dimension: a build that
+// predates them (prod and dev share one DB) keeps writing its rows unchanged,
+// and an existing row's totals never split, so nothing is counted twice.
+const LONG_TOKEN_COLS = [
+  "long_input_tokens",
+  "long_output_tokens",
+  "long_cache_read_tokens",
+  "long_cache_write_tokens",
+  "long_cache_write_1h_tokens",
+];
+for (const table of ["token_usage", "token_usage_hourly"]) {
+  try {
+    db.prepare(`SELECT long_input_tokens FROM ${table} LIMIT 1`).get();
+  } catch {
+    for (const col of LONG_TOKEN_COLS) {
+      db.prepare(`ALTER TABLE ${table} ADD COLUMN ${col} INTEGER NOT NULL DEFAULT 0`).run();
+    }
+  }
+}
+
 // Startup cleanup: mark stale active sessions as completed.
 // Legacy sessions (created before SessionEnd hook) will never receive a SessionEnd event,
 // so they stay "active" forever. Complete any active session whose last event is older than
@@ -2052,6 +2107,38 @@ const stmts = {
        OR excluded.web_fetch_requests > web_fetch_requests
        OR excluded.code_execution_requests > code_execution_requests
   `),
+  // A bucket's long-context subsets (see the token_usage columns), written
+  // after its row by the same caller. Args: the five long_* values in column
+  // order, then the row's key. RAISE keeps the high-water mark (the live and
+  // import writers, like the hourly rows); SET is the exact reconcile path's.
+  raiseLongTokens: db.prepare(`
+    UPDATE token_usage SET
+      long_input_tokens = MAX(long_input_tokens, ?),
+      long_output_tokens = MAX(long_output_tokens, ?),
+      long_cache_read_tokens = MAX(long_cache_read_tokens, ?),
+      long_cache_write_tokens = MAX(long_cache_write_tokens, ?),
+      long_cache_write_1h_tokens = MAX(long_cache_write_1h_tokens, ?)
+    WHERE session_id = ? AND model = ? AND speed = ? AND inference_geo = ? AND service_tier = ?
+  `),
+  setLongTokens: db.prepare(`
+    UPDATE token_usage SET
+      long_input_tokens = ?,
+      long_output_tokens = ?,
+      long_cache_read_tokens = ?,
+      long_cache_write_tokens = ?,
+      long_cache_write_1h_tokens = ?
+    WHERE session_id = ? AND model = ? AND speed = ? AND inference_geo = ? AND service_tier = ?
+  `),
+  raiseLongHourly: db.prepare(`
+    UPDATE token_usage_hourly SET
+      long_input_tokens = MAX(long_input_tokens, ?),
+      long_output_tokens = MAX(long_output_tokens, ?),
+      long_cache_read_tokens = MAX(long_cache_read_tokens, ?),
+      long_cache_write_tokens = MAX(long_cache_write_tokens, ?),
+      long_cache_write_1h_tokens = MAX(long_cache_write_1h_tokens, ?)
+    WHERE session_id = ? AND hour = ? AND model = ? AND speed = ? AND inference_geo = ?
+      AND service_tier = ?
+  `),
   getTokenTotals: db.prepare(`
     SELECT
       COALESCE(SUM(input_tokens + baseline_input), 0) as total_input,
@@ -2073,7 +2160,9 @@ const stmts = {
       cache_write_1h_tokens + baseline_cache_write_1h as cache_write_1h_tokens,
       web_search_requests + baseline_web_search as web_search_requests,
       web_fetch_requests + baseline_web_fetch as web_fetch_requests,
-      code_execution_requests + baseline_code_execution as code_execution_requests
+      code_execution_requests + baseline_code_execution as code_execution_requests,
+      long_input_tokens, long_output_tokens, long_cache_read_tokens,
+      long_cache_write_tokens, long_cache_write_1h_tokens
     FROM token_usage WHERE session_id = ?`
   ),