examples: full marketing-agent demo with streaming + multi-turn + file output

Runs github.com/shreyas-lyzr/marketing-agent (32-skill GAP agent) via
ComputerAgent with LocalSubstrate. Demonstrates:
- Multi-turn conversation (product context → cold email → pricing → launch)
- Real-time streaming of assistant text and tool calls
- File-backed SessionStore for cross-process resume
- Harness FS API to fetch agent-produced files
- `await using` for clean dispose

Usage:
  ANTHROPIC_API_KEY=sk-... bun run examples/marketing-agent.ts
  ANTHROPIC_API_KEY=sk-... bun run examples/marketing-agent.ts resume

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Author: kapaleshreyas

examples/marketing-agent.ts

@@ -0,0 +1,279 @@
+/**
+ * Full-fledged marketing agent example.
+ *
+ * Runs the marketing-agent GAP repo (github.com/shreyas-lyzr/marketing-agent)
+ * as a multi-turn session: first turn sets product context, subsequent turns
+ * request specific marketing deliverables. Each response streams in real-time
+ * and the final outputs are saved to markdown files under ./marketing-outputs/.
+ *
+ * Usage:
+ *   ANTHROPIC_API_KEY=sk-ant-... bun run examples/marketing-agent.ts
+ *
+ * On subsequent runs, the session is resumed from disk so the agent remembers
+ * the product context set in the first turn.
+ *
+ *   ANTHROPIC_API_KEY=sk-ant-... bun run examples/marketing-agent.ts resume
+ *
+ * What this demonstrates:
+ *   - Running a real public GAP repo via ComputerAgent
+ *   - Streaming every sdk_message and tool call as they arrive
+ *   - Multi-turn conversation with a file-backed SessionStore
+ *   - Fetching agent-produced files via the Harness FS API
+ *   - clean dispose() via `await using`
+ */
+
+import { mkdir, writeFile, readFile } from "node:fs/promises";
+import { existsSync } from "node:fs";
+import { join } from "node:path";
+import { ComputerAgent, LocalSubstrate } from "computeragent";
+
+const ANTHROPIC_API_KEY = process.env.ANTHROPIC_API_KEY;
+if (!ANTHROPIC_API_KEY) {
+  console.error("Error: ANTHROPIC_API_KEY is not set.");
+  console.error("  export ANTHROPIC_API_KEY=sk-ant-...");
+  process.exit(1);
+}
+
+const OUTPUTS_DIR = join(import.meta.dir ?? __dirname, "marketing-outputs");
+const SESSION_FILE = join(OUTPUTS_DIR, ".session-id");
+const RESUME = process.argv[2] === "resume";
+
+// ── Helpers ──────────────────────────────────────────────────────────────────
+
+function hr(label: string): void {
+  const line = "─".repeat(60);
+  console.log(`\n${line}`);
+  console.log(`  ${label}`);
+  console.log(`${line}`);
+}
+
+function dim(s: string): string {
+  return `\x1b[2m${s}\x1b[0m`;
+}
+
+function green(s: string): string {
+  return `\x1b[32m${s}\x1b[0m`;
+}
+
+function yellow(s: string): string {
+  return `\x1b[33m${s}\x1b[0m`;
+}
+
+async function streamTurn(
+  agent: ComputerAgent,
+  message: string,
+): Promise<{ sessionId: string; result: string; harnessUrl: string }> {
+  console.log(`\n${yellow("You:")} ${message}\n`);
+
+  let result = "";
+  let sessionId = "";
+
+  const handle = agent.chat(message);
+
+  for await (const ev of handle) {
+    if (ev.kind === "ca_session_started") {
+      sessionId = ev.sessionId;
+      console.log(dim(`[session ${sessionId} • engine=${ev.engine}]`));
+    } else if (ev.kind === "sdk_message") {
+      const p = ev.payload as Record<string, unknown>;
+
+      if (p.type === "assistant") {
+        const msg = p.message as { content?: unknown[] } | undefined;
+        for (const block of msg?.content ?? []) {
+          const b = block as { type?: string; text?: string; name?: string; input?: unknown };
+          if (b.type === "text" && b.text) {
+            // Stream text progressively — print in chunks
+            process.stdout.write(b.text);
+          } else if (b.type === "tool_use") {
+            const input = JSON.stringify(b.input ?? {});
+            console.log(`\n${dim(`  → ${b.name}(${input.slice(0, 120)}${input.length > 120 ? "…" : ""})`)}`);
+          }
+        }
+      } else if (p.type === "tool") {
+        // tool result
+        const content = typeof p.content === "string" ? p.content : JSON.stringify(p.content);
+        console.log(dim(`    ← ${content.slice(0, 200)}${content.length > 200 ? "…" : ""}`));
+      } else if (p.type === "result" && typeof p.result === "string") {
+        result = p.result;
+      }
+    } else if (ev.kind === "ca_usage_snapshot") {
+      const u = ev as unknown as { input_tokens?: number; output_tokens?: number; cost_usd?: number };
+      if (u.cost_usd !== undefined) {
+        console.log(`\n${dim(`[usage: in=${u.input_tokens} out=${u.output_tokens} cost=$${u.cost_usd?.toFixed(4)}]`)}`);
+      }
+    } else if (ev.kind === "ca_session_ended") {
+      console.log(`\n${dim(`[ended: ${ev.reason}]`)}`);
+    }
+  }
+
+  const harnessUrl = await agent.harnessUrl();
+  return { sessionId, result, harnessUrl };
+}
+
+async function saveOutput(filename: string, content: string): Promise<void> {
+  const path = join(OUTPUTS_DIR, filename);
+  await writeFile(path, content, "utf8");
+  console.log(`\n${green("✓")} Saved → ${path}`);
+}
+
+async function fetchAgentFile(harnessUrl: string, sessionId: string, path: string): Promise<string | null> {
+  try {
+    const res = await fetch(`${harnessUrl}/v1/sessions/${sessionId}/fs/file?path=${encodeURIComponent(path)}`);
+    if (!res.ok) return null;
+    return await res.text();
+  } catch {
+    return null;
+  }
+}
+
+// ── Main ─────────────────────────────────────────────────────────────────────
+
+await mkdir(OUTPUTS_DIR, { recursive: true });
+
+let priorSessionId: string | undefined;
+if (RESUME && existsSync(SESSION_FILE)) {
+  priorSessionId = (await readFile(SESSION_FILE, "utf8")).trim();
+  console.log(`\nResuming session: ${priorSessionId}`);
+}
+
+// Boot the marketing agent from the public GitHub repo.
+// LocalSubstrate spawns the harness in a subprocess — no Docker required.
+await using agent = new ComputerAgent({
+  source: {
+    type: "git",
+    url: "github.com/shreyas-lyzr/marketing-agent",
+  },
+  harness: "claude-agent-sdk",
+  runtime: new LocalSubstrate(),
+  envs: { ANTHROPIC_API_KEY },
+  options: {
+    permissionMode: "bypassPermissions",
+    settingSources: ["project"],
+  },
+  sessionStore: {
+    kind: "file",
+    options: { root: OUTPUTS_DIR },
+  },
+  ...(priorSessionId ? { sessionId: priorSessionId } : {}),
+});
+
+// ── Turn 1: Product context ───────────────────────────────────────────────────
+
+if (!RESUME) {
+  hr("Turn 1 — Product Context");
+
+  const { sessionId } = await streamTurn(
+    agent,
+    `Here is our product context. Please confirm you've loaded it and identify the top 3
+marketing challenges you'd recommend we tackle first.
+
+Product: Lyzr AI — an enterprise platform for building, deploying, and governing AI agents at scale.
+Audience: Mid-market and enterprise engineering and IT leaders (CTOs, VP Engineering, Head of AI).
+Stage: Series A, $8M ARR, 120 customers, growing 15% MoM.
+Key differentiators: On-prem / VPC deployment, SOC2 compliant, LLM-agnostic (OpenAI, Anthropic, Gemini, Mistral).
+Main competitors: LangChain, CrewAI, Vertex AI Agent Builder.
+Current channels: mostly outbound sales, some inbound content, limited PLG motion.
+Top conversion barrier: enterprises want a PoC before committing — long evaluation cycles (60-90 days).`,
+  );
+
+  await writeFile(SESSION_FILE, sessionId, "utf8");
+  console.log(`\nSession ID saved → ${SESSION_FILE}`);
+}
+
+// ── Turn 2: Cold email sequence ───────────────────────────────────────────────
+
+hr("Turn 2 — Cold Email Sequence");
+
+const { result: emailResult } = await streamTurn(
+  agent,
+  `Write a 4-email cold outreach sequence targeting VP Engineering at Series B+ SaaS companies
+(500–2000 employees). The goal of the sequence is to book a 30-minute PoC scoping call.
+Angle: they are likely already using LangChain or CrewAI and hitting reliability or governance issues at scale.
+Format each email with: Subject line, Body, Send timing.`,
+);
+
+await saveOutput(
+  "cold-email-sequence.md",
+  `# Cold Email Sequence — VP Engineering @ Series B+ SaaS\n\nGenerated by marketing-agent via ComputerAgent\n\n---\n\n${emailResult}`,
+);
+
+// ── Turn 3: Pricing strategy ──────────────────────────────────────────────────
+
+hr("Turn 3 — Pricing Strategy");
+
+const { result: pricingResult } = await streamTurn(
+  agent,
+  `Design a 3-tier pricing structure for Lyzr AI. We currently charge flat enterprise contracts
+($3k–$15k/month). We want to add a self-serve tier to capture PLG motion and reduce the
+60-90 day sales cycle for smaller deals. Include:
+- Tier names and positioning
+- Pricing metric recommendation (per-seat vs. per-agent vs. usage-based)
+- Feature gates between tiers
+- Recommended price points with reasoning
+- A/B test ideas for the pricing page`,
+);
+
+await saveOutput(
+  "pricing-strategy.md",
+  `# Pricing Strategy — Lyzr AI\n\nGenerated by marketing-agent via ComputerAgent\n\n---\n\n${pricingResult}`,
+);
+
+// ── Turn 4: Launch strategy for self-serve tier ───────────────────────────────
+
+hr("Turn 4 — Self-Serve Launch Strategy");
+
+const { result: launchResult, sessionId: finalSession, harnessUrl } = await streamTurn(
+  agent,
+  `Now create a 30-day launch plan for the new self-serve tier.
+We want 500 signups in the first 30 days. Channels available: Product Hunt, HN Show,
+our existing email list (4,200 subscribers), LinkedIn (8k followers), and developer communities.
+Include: Pre-launch checklist, launch day playbook, week-by-week activation plan,
+success metrics, and the top 3 risks with mitigations.`,
+);
+
+await saveOutput(
+  "launch-strategy.md",
+  `# Self-Serve Launch Strategy — 30-Day Plan\n\nGenerated by marketing-agent via ComputerAgent\n\n---\n\n${launchResult}`,
+);
+
+// ── Bonus: check if agent wrote any files to its workdir ─────────────────────
+
+hr("Harness Filesystem");
+
+try {
+  const treeRes = await fetch(
+    `${harnessUrl}/v1/sessions/${finalSession}/fs/tree?depth=2`,
+  );
+  if (treeRes.ok) {
+    const tree = (await treeRes.json()) as { entries: { path: string; type: string; size: number }[] };
+    const interesting = tree.entries.filter((e) => e.path !== "/" && !e.path.startsWith("/."));
+    if (interesting.length > 0) {
+      console.log("\nFiles the agent wrote to its workspace:");
+      for (const e of interesting) {
+        console.log(`  ${e.type.padEnd(4)} ${e.path.padEnd(50)} ${e.size}b`);
+        if (e.type === "file") {
+          const content = await fetchAgentFile(harnessUrl, finalSession, e.path);
+          if (content) {
+            const outName = e.path.replace(/^\//, "").replace(/\//g, "-");
+            await saveOutput(`agent-workdir-${outName}`, content);
+          }
+        }
+      }
+    } else {
+      console.log(dim("  (agent workdir is empty — all output was inline text)"));
+    }
+  }
+} catch {
+  // FS API optional — harness may not expose it in all configurations
+}
+
+// ── Summary ───────────────────────────────────────────────────────────────────
+
+hr("Done");
+console.log(`\nAll outputs saved to: ${OUTPUTS_DIR}/`);
+console.log("  cold-email-sequence.md   — 4-email VP Engineering outreach sequence");
+console.log("  pricing-strategy.md      — 3-tier SaaS pricing with feature gates");
+console.log("  launch-strategy.md       — 30-day self-serve launch plan");
+console.log(`\nSession ID: ${finalSession}`);
+console.log("Run with 'resume' to continue this session in a fresh process:");
+console.log(`  ANTHROPIC_API_KEY=sk-... bun run examples/marketing-agent.ts resume\n`);

examples/package.json

@@ -4,9 +4,11 @@
   "private": true,
   "type": "module",
   "scripts": {
-    "wedge1": "bun run wedge1-server.ts"
+    "wedge1": "bun run wedge1-server.ts",
+    "marketing-agent": "bun run marketing-agent.ts"
   },
   "dependencies": {
+    "computeragent": "workspace:*",
     "@computeragent/engine-claude-agent-sdk": "workspace:*",
     "@computeragent/engine-gitagent": "workspace:*",
     "@computeragent/harness-server": "workspace:*",