great-cto 2.10.2 → 2.11.0

package/dist/mcp.js

@@ -126,6 +126,112 @@ function toolQueryDecisions(args) {
         }),
     };
 }
+// ── Board tools — call the running board HTTP API ─────────────────────────
+// Board port: $GREAT_CTO_PORT (default 3141).
+const BOARD_PORT = parseInt(process.env.GREAT_CTO_PORT ?? "3141", 10);
+const BOARD_BASE = `http://127.0.0.1:${BOARD_PORT}`;
+async function boardFetch(path) {
+    try {
+        const res = await fetch(`${BOARD_BASE}${path}`);
+        if (!res.ok)
+            throw new Error(`HTTP ${res.status}`);
+        return res.json();
+    }
+    catch (e) {
+        throw new Error(`Board unreachable at ${BOARD_BASE} — run \`great-cto board\` first. (${e.message})`);
+    }
+}
+function boardPqs(project) {
+    return project ? `?project=${encodeURIComponent(project)}` : "";
+}
+async function toolProjectStatus(args) {
+    const qs = boardPqs(args.project);
+    const [inbox, metrics] = await Promise.all([
+        boardFetch(`/api/inbox${qs}`),
+        boardFetch(`/api/metrics${qs}`),
+    ]);
+    const s = inbox.summary ?? {};
+    const lines = [
+        `## Project status${args.project ? ` — ${args.project}` : ""}`,
+        "",
+        `**Open gates:** ${s.gates ?? 0}`,
+        `**Blocked tasks:** ${s.blocked ?? 0}`,
+        `**P0 incidents:** ${s.p0 ?? 0}`,
+        `**Stale in-progress:** ${s.stale ?? 0}`,
+    ];
+    if ((inbox.pending_gates ?? []).length > 0) {
+        lines.push("", "### Gates awaiting approval");
+        for (const g of inbox.pending_gates.slice(0, 5)) {
+            lines.push(`- **${g.id}** ${g.title} _(${g.status})_`);
+        }
+    }
+    if ((inbox.blocked ?? []).length > 0) {
+        lines.push("", "### Blocked tasks");
+        for (const b of inbox.blocked.slice(0, 5)) {
+            lines.push(`- **${b.id}** ${b.title}`);
+        }
+    }
+    if (metrics) {
+        lines.push("", `**Tasks:** ${metrics.done ?? 0}/${metrics.total ?? 0} done`);
+    }
+    return lines.join("\n");
+}
+async function toolCostSummary(args) {
+    const days = Math.min(365, Math.max(1, args.days ?? 30));
+    const qs = boardPqs(args.project);
+    const d = await boardFetch(`/api/cost${qs}${qs ? "&" : "?"}days=${days}`);
+    const lines = [
+        `## Cost summary${args.project ? ` — ${args.project}` : ""} (last ${days} days)`,
+        "",
+        `**Total LLM spend:** $${(d.total_llm ?? 0).toFixed(2)}`,
+        `**Daily burn:** $${(d.daily_avg ?? 0).toFixed(2)}/day`,
+        `**Projected monthly:** $${(d.projected_monthly ?? 0).toFixed(0)}`,
+    ];
+    if (d.monthly_budget) {
+        const status = d.over_budget ? "⚠️ OVER BUDGET" : "✅ within budget";
+        lines.push(`**Budget:** $${d.monthly_budget}/month — ${status}`);
+    }
+    if (d.savings_x)
+        lines.push(`**vs Human team:** ${d.savings_x}× cheaper`);
+    if ((d.by_feature ?? []).length > 0) {
+        lines.push("", "### Top features by AI spend");
+        for (const f of d.by_feature.slice(0, 8)) {
+            lines.push(`- **${f.feature}**: $${f.llm.toFixed(2)} (${f.runs} runs)`);
+        }
+    }
+    return lines.join("\n");
+}
+async function toolPipelineStages(args) {
+    const stages = await boardFetch(`/api/pipeline${boardPqs(args.project)}`);
+    const lines = [
+        `## Pipeline stages${args.project ? ` — ${args.project}` : ""}`,
+        "",
+        "| Stage | Status | Verdict | Last run |",
+        "|---|---|---|---|",
+    ];
+    for (const s of stages ?? []) {
+        lines.push(`| ${s.stage} | ${s.status ?? "idle"} | ${s.verdict ?? "—"} | ${(s.last_ts ?? "—").slice(0, 16)} |`);
+    }
+    return lines.join("\n");
+}
+async function toolRecentVerdicts(args) {
+    const limit = Math.min(50, Math.max(1, args.limit ?? 10));
+    const data = await boardFetch(`/api/metrics${boardPqs(args.project)}`);
+    const verdicts = (data.verdicts ?? []).slice(-limit).reverse();
+    if (verdicts.length === 0)
+        return "No recent verdicts found.";
+    const lines = [
+        `## Recent verdicts${args.project ? ` — ${args.project}` : ""} (last ${verdicts.length})`,
+        "",
+        "| Time | Agent | Verdict | Cost |",
+        "|---|---|---|---|",
+    ];
+    for (const v of verdicts) {
+        const cost = v.cost_usd != null ? `$${v.cost_usd.toFixed(2)}` : "—";
+        lines.push(`| ${(v.ts ?? "").slice(0, 16)} | ${v.agent ?? "—"} | ${v.verdict ?? "—"} | ${cost} |`);
+    }
+    return lines.join("\n");
+}
 // ── Tool dispatch table ────────────────────────────────────────────────────
 const TOOLS = [
     {
@@ -194,6 +300,57 @@ const TOOLS = [
         },
         handler: toolQueryDecisions,
     },
+    // ── Board tools (require running board at $GREAT_CTO_PORT / 3141) ────────
+    {
+        name: "project_status",
+        description: "Get current pipeline status from the great_cto board: open gates awaiting approval, blocked tasks, P0 incidents, and in-progress work. " +
+            "Use this before starting expensive agent tasks to check for blockers. Requires `great-cto board` running.",
+        inputSchema: {
+            type: "object",
+            properties: {
+                project: { type: "string", description: "Project slug (optional, defaults to board's active project)" },
+            },
+        },
+        handler: toolProjectStatus,
+    },
+    {
+        name: "cost_summary",
+        description: "Get AI agent LLM spend from the board: total cost, daily burn rate, projected monthly cost, budget status, " +
+            "and cost broken down by feature. Use to check budget before spawning resource-intensive agents. Requires `great-cto board` running.",
+        inputSchema: {
+            type: "object",
+            properties: {
+                days: { type: "number", description: "Window in days: 1, 7, 30, 90, 365 (default: 30)" },
+                project: { type: "string", description: "Project slug (optional)" },
+            },
+        },
+        handler: toolCostSummary,
+    },
+    {
+        name: "pipeline_stages",
+        description: "Get full pipeline stage list with status (idle/done/failed) and last agent verdict. " +
+            "Shows which stage the current feature is at and what each agent last decided. Requires `great-cto board` running.",
+        inputSchema: {
+            type: "object",
+            properties: {
+                project: { type: "string", description: "Project slug (optional)" },
+            },
+        },
+        handler: toolPipelineStages,
+    },
+    {
+        name: "recent_verdicts",
+        description: "Get the most recent agent verdict lines from the board: timestamps, agent names, verdict values (APPROVED/DONE/BLOCKED), and costs. " +
+            "Quick recap of what agents have done recently. Requires `great-cto board` running.",
+        inputSchema: {
+            type: "object",
+            properties: {
+                limit: { type: "number", description: "Number of verdicts to return (1–50, default: 10)" },
+                project: { type: "string", description: "Project slug (optional)" },
+            },
+        },
+        handler: toolRecentVerdicts,
+    },
 ];
 // ── JSON-RPC handler ───────────────────────────────────────────────────────
 async function handle(req) {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "great-cto",
-  "version": "2.10.2",
+  "version": "2.11.0",
   "description": "One command install for the great_cto Claude Code plugin. Auto-detects your stack, picks the right archetype, bootstraps PROJECT.md.",
   "keywords": [
     "claude-code",