# Harshit Jain — PMM + AI builder > Principal PMM at ClickUp with 10 years in product marketing. Uses Claude Code, Codex, and Cursor to ship the internal tools marketing teams normally have to ask engineering for — competitive intelligence CLIs, content pipelines, agent routers, memory layers for AI coding tools. Writes the working version of all of it in QBack, a weekly newsletter. Source: https://qback.ai Generated: 2026-09-04T16:50:36.228Z This document contains the complete content of qback.ai. For live/structured access, this site also runs an MCP server at https://qback.ai/mcp and exposes WebMCP tools in-page via `document.modelContext`. --- # Profile - **Name:** Harshit Jain - **Role:** Principal Product Marketing Manager at ClickUp - **Experience:** 10+ years in product marketing - **Companies:** ClickUp, Observe.AI, LeadSquared, QBack, e6data - **Categories marketed:** conversational intelligence, CRM, sales enablement, data infrastructure, work management - **Skills:** product marketing, positioning and messaging, competitive intelligence, go-to-market strategy, product launches, sales enablement, AI agent workflows, Claude Code, Codex, Cursor Principal PMM at ClickUp. Before that, usually the first PMM hire at places like Observe.AI, LeadSquared, QBack, and e6data. Different categories, same job: launch products, sharpen messaging, run competitive intel, and make the thing easier to sell. That also means marketing to everyone from very technical buyers to teams that basically live in spreadsheets. AI rewrote the playbook, so instead of just talking about it he started building with Claude Code, Codex and Cursor and writing the working version down in public on QBack. ## Stats - **10+** — Years GTM - **42** — Newsletter editions - **3** — AI agents in the stack - **5** — Companies ## Links - X (@harshitqback): https://twitter.com/harshitqback - LinkedIn (harshitjain7): https://www.linkedin.com/in/harshitjain7/ - GitHub (qb-harshit): https://github.com/qb-harshit - Newsletter (QBack): https://newsletter.qback.ai _No public email by design. Best contact is a DM on X/Twitter or a LinkedIn connection._ --- # Pages ## https://qback.ai/ # Harshit Jain — PMM + AI builder > I build the tools PMMs keep asking engineering for. Principal PMM at ClickUp. I use Claude Code / Codex / Cursor to ship what marketing teams usually can't. ## About 10 years of product marketing. Now I build with AI agents from my terminal. Principal PMM at ClickUp. Before that, I was usually the first PMM hire at places like Observe.AI, LeadSquared, QBack, and e6data. Different categories, same job: launch products, sharpen messaging, run competitive intel, and make the thing easier to sell. That also means I've marketed to everyone from very technical buyers to teams that basically live in spreadsheets. AI rewrote the playbook, so instead of just talking about it I started building with Claude Code, Codex and Cursor and writing the working version down in public on QBack. ## Selected builds Stuff I built because waiting was getting annoying. - **QBack Newsletter** — weekly playbooks on using Claude Code and Cursor for PMM work. Real workflows, no AI fluff. https://newsletter.qback.ai - **Competitive Intel CLI** — I got tired of bloated intel tools and generic prompts, so I built a CLI that tracks what actually changed. https://github.com/qb-harshit/competitive-intel-cli - **fix-claude-mem** — a Claude Code skill that gives sessions a memory layer. Digests write themselves, local search indexes everything, next session picks up where the last one left off. https://qback.ai/fix-claude-mem ## Newsletter QBack is where I dump the playbooks, teardowns, and "I built this last night" experiments behind the work on this site. Weekly, at https://newsletter.qback.ai ## Contact Send me the messy one. You can find me on X (@harshitqback), LinkedIn (harshitjain7), or GitHub (qb-harshit). There's no public email on the site on purpose — DMs work better anyway. --- ## https://qback.ai/fix-claude-mem # fix-claude-mem **A Claude Code skill · MIT licensed · fully local** > Claude Code keeps forgetting everything. This fixes it. The way it works is pretty simple — when you close the terminal, a hook writes down what just happened in `memory/sessions/`. qmd indexes that locally. Next time you open Claude Code, it pulls the most relevant past memories into context before you even type anything. So Claude just knows what you worked on last week instead of asking. ## The problem **You explain your codebase to Claude, and an hour later it's gone.** You walk Claude through your architecture, point at the files that matter, work through a nasty bug, land on a good pattern. Then you close the terminal and it's gone. Next session you're explaining it all over again. So you start pasting the same context into every new session. Your `claude.md` grows from 100 lines to 1000. Eventually you stop bothering with long sessions because most of it evaporates anyway. The actual bug isn't that Claude forgets. It's that there's no memory layer sitting between you and Claude. Each session literally starts at zero, and three months in you're still re-explaining the same things you explained on day one. What breaks: - Context vanishes when the terminal closes - `claude.md` becomes a 2000-line dumping ground - You re-derive decisions you already made - Pasting the same context into every new session - Three months of work doesn't compound into anything ## How it works Three hooks, one local index, memory that compounds. **01 — Session ends.** A stop hook fires when you close the terminal. It writes a short digest to `memory/sessions/` — what you built, what you decided, what's still open. Plain markdown with frontmatter so qmd can index it. **02 — qmd indexes everything.** qmd (by @tobi) runs locally and indexes everything in your memory directory. It does hybrid search — keyword, vectors, reranking — using three small models that live on your machine. Nothing leaves your laptop. **03 — Session starts.** Next time you open Claude Code, a hook reads your first message, queries qmd for the most relevant past memories, and injects them as context. So Claude wakes up already knowing what you worked on. ## Install One command if you use Claude Code. Plugin marketplace: ``` /plugin marketplace add qb-harshit/fix-claude-mem ``` Then run setup in your project: ``` fix-claude-mem setup ``` Or clone the repo manually: ``` git clone https://github.com/qb-harshit/fix-claude-mem ~/.claude/skills/fix-claude-mem ``` Setup makes the memory directories at your project root, wires the three hooks into `.claude/settings.json` without touching what's already there, and installs qmd if you don't have it yet. The whole thing takes about 30 seconds. First run also pulls qmd's models down which is around 2GB and another couple of minutes. After that it just runs in the background and you forget about it. ## For product marketers If you're a PMM using Claude Code, you have a memory problem bigger than engineers do. Engineering has git to remember everything. PMM work doesn't have that. Your research, your objection patterns, your voice rules, your competitive intel — it's all in your head or scattered across Notion and Slack threads nobody can find. So every Claude session starts from scratch. **Competitive intel.** Before: you ran a scrape last week, noticed a positioning shift, flagged a pricing change. Close the terminal and it's basically gone. After: it all stays searchable. Next session you just ask "what changed with competitor X last month" and Claude has the answer without you pasting anything. **Buyer research.** Before: call summaries, quotes, objection patterns — you paste the same ones into every positioning session. After: call notes pile up in `memory/decisions/` over months. Your next positioning exercise starts with three months of real objection data already loaded. **Positioning docs.** Before: your messaging framework gets rewritten every month because nobody remembers the decisions behind the current one. After: every positioning decision gets captured with the reasoning attached. The framework evolves across quarters instead of resetting every time the market shifts. **Content pipeline.** Before: Claude writes generic AI-slop because it has no idea about your voice rules, your banned words, or the specific phrases you actually use. After: voice rules live in `memory/decisions/`. Every content session starts with those already in context. The pattern that works in practice: one memory directory per area — `intel/`, `research/`, `positioning/`, `content/` — with a short `claude.md` at each root explaining what that area is about. Point fix-claude-mem at them and let it do the rest. ## Compared to other ways to give Claude memory | | fix-claude-mem | manual claude.md | cursor chat history | mcp-memory | |---|---|---|---|---| | survives session restart | yes | yes | no | yes | | compounds over time | yes | no | no | partial | | hybrid search (keyword + vector + rerank) | yes | no | no | keyword only | | works fully offline | yes | yes | no | depends | | auto-generated digests | yes | no | yes | partial | | cross-project search | yes | no | no | yes | | runs on laptop | yes | yes | n/a | yes | | zero lock-in (plain markdown) | yes | yes | no | no | ## License and credit MIT licensed, all the code is open. Fork it, ship your own version, do whatever you want with it. Runs on [qmd](https://github.com/tobi/qmd) by [@tobi](https://x.com/tobi). Repo: https://github.com/qb-harshit/fix-claude-mem --- # Builds ## QBack Newsletter - **Kind:** live publication - **Status:** live - **Tech:** Python, Astro, Remotion - **Topics:** newsletter, content, ai, marketing, claude code - **Link:** https://newsletter.qback.ai Weekly playbooks on using Claude Code and Cursor for PMM work. Real workflows, no AI fluff. A weekly newsletter about AI, marketing, and building in public, produced with a multi-agent pipeline: voice capture, AI drafting, image generation, and video rendering. It is where the long version of everything else on this site gets written down. ## Competitive Intel CLI - **Kind:** recent build - **Status:** open-source - **Tech:** Python, Click, BeautifulSoup - **Topics:** competitive intelligence, cli, python, pmm tooling - **Link:** https://github.com/qb-harshit/competitive-intel-cli - **Write-up:** https://newsletter.qback.ai/p/why-pay-40k-for-competitive-intel Got tired of bloated intel tools and generic prompts, so built a CLI that tracks what actually changed. A command-line tool that scrapes, analyzes, and reports on competitor positioning, pricing, and feature sets. Built for PMMs who think in terminals — the point is diffing what actually changed since last week, not generating another 40-page report nobody reads. ## fix-claude-mem - **Kind:** open source skill - **Status:** open-source - **Tech:** Claude Code, qmd, Bash, Markdown - **Topics:** claude code, memory, developer tools, open source, rag - **Link:** https://github.com/qb-harshit/fix-claude-mem - **Write-up:** https://qback.ai/fix-claude-mem A Claude Code skill that gives sessions a memory layer. Digests write themselves, local search indexes everything, the next session picks up where the last one left off. Three hooks and one local index. A stop hook writes a session digest to memory/sessions/ when you close the terminal. qmd indexes it locally with hybrid search. A session-start hook reads your first message, queries qmd, and injects the most relevant past memories as context. MIT licensed, fully local, nothing leaves the laptop. ## Newsletter Image Generator - **Kind:** tool - **Status:** open-source - **Tech:** Python, Pillow, NumPy - **Topics:** image generation, python, design, newsletter - **Link:** https://github.com/qb-harshit/qback-write-like-harshit Retro dithered header image generator for the newsletter. A Python tool that generates CRT-styled, dithered newsletter headers with unicode block ASCII art. 13 color schemes, multiple dither modes, chromatic aberration effects. ## Video Pipeline - **Kind:** tool - **Status:** internal - **Tech:** TypeScript, Remotion, React - **Topics:** video, remotion, content pipeline, automation - **Link:** https://newsletter.qback.ai Remotion-based newsletter-to-video pipeline. Converts newsletter content into CRT-aesthetic announcement videos. Full pipeline: content parsing, scene generation, overlay effects (scanlines, grain, chromatic aberration), and Remotion rendering. ## Agent Router - **Kind:** tool - **Status:** internal - **Tech:** Python, Claude API - **Topics:** ai agents, routing, developer tools, python - **Link:** https://newsletter.qback.ai Smart router across three AI coding agents. Routes prompts to the optimal AI coding agent (Amp, Claude Code, or Cursor) based on intent classification, credit budgets, and task complexity. Voice-to-text friendly, because every prompt into it is dictated. --- # QBack newsletter Weekly playbooks on using Claude Code and Cursor for product marketing work. Real workflows, teardowns, and "I built this last night" experiments — no AI fluff. - **URL:** https://newsletter.qback.ai - **Feed:** https://newsletter.qback.ai/feed - **Cadence:** weekly - **Audience:** product marketers, GTM operators, and PMMs who want to build instead of file tickets ## Archive ### How to build images and videos inside Cursor and Claude Code using Remotion Feb 3, 2026 — https://newsletter.qback.ai/p/build-images-videos-cursor-claude-code-remotion I talk to my editor, it makes images and videos. that’s it. ### How to rank content on AI search engines (ChatGPT, Claude, Perplexity, Grok, Gemini) Oct 17, 2025 — https://newsletter.qback.ai/p/how-to-rank-content-on-ai-search-engines-chatgpt-claude-perplexity-grok content growth hacks to become source of truth for LLMs ### How to become the marketer AI can't replace: build, ship, position Oct 1, 2025 — https://newsletter.qback.ai/p/how-to-become-ai-proof-marketer-future-guide-for-pmm AI-first thinking + context trinity + ship hyper-specific stuff at scale = unfair advantage ### Your product is your only positioning Sep 19, 2025 — https://newsletter.qback.ai/p/your-product-is-your-only-ai-positioning hide the plumbing, show the fountain. sharing a few hacks and examples to do it ### Why pay $40k for competitive intel tools when you can build it for free with AI? Sep 11, 2025 — https://newsletter.qback.ai/p/why-pay-40k-for-competitive-intel how I built this entirely inside Cursor that actually works and updates whenever I want (video tutorial + GitHub link added for DIY) ### I built a Cursor playbook that writes my newsletter and all content exactly like me Aug 25, 2025 — https://newsletter.qback.ai/p/ai-marketing-content-generator-cursor Send this URL to Cursor/Claude and ask it to build this for you. It takes raw thoughts and ships newsletters, LinkedIn posts, and tweets in your own voice. ### I built (vibe-coded) an AI agent for NPS analysis Jul 2, 2025 — https://newsletter.qback.ai/p/how-to-build-vibe-code-an-ai-agent-for-nps-analysis-diy-for-product-marketers As a PMM, if you’re tracking the user sentiment and voice of customer analysis, then you can built this into an in-house app with custom logic and prompts specific to your company process. ### ABM That Doesn’t Suck AI Edition (Clay × Webflow × Zapier) Jun 23, 2025 — https://newsletter.qback.ai/p/abm-that-doesnt-suck-ai-edition-clay because “Hey {first_name}” doesn’t cut it anymore. ### Built a Dev Homepage Without Burning $10K on Random Agency BS Apr 13, 2025 — https://newsletter.qback.ai/p/built-a-dev-homepage-without-burning About to blow $10K on some agency to write your homepage? Stop! Unless you’ve got zero product marketing chops, write it yourself. Devs don’t want your story, they want to know if your product works. ### Announcing QBack: The AI Quarterback for GTM Mar 20, 2025 — https://newsletter.qback.ai/p/announcing-qback-newsletter-inside From Harshit Jain --- # Agent interfaces ## Remote MCP server Endpoint: `https://qback.ai/mcp` (Model Context Protocol, Streamable HTTP transport, no authentication). Tools: `search_site`, `get_profile`, `list_builds`, `get_build`, `get_page`, `list_newsletter_posts`, `get_contact_channels`. Add it in Claude, Claude Code, or any MCP client as a remote/custom connector pointed at that URL. ## WebMCP Every page registers the same tools in-page through `document.modelContext`, so a browser-resident agent can call them directly on the live DOM. The homepage additionally exposes page-control tools (scroll to a section, drive the on-page terminal). ## Static endpoints - `https://qback.ai/llms.txt` — short index - `https://qback.ai/llms-full.txt` — this document - `https://qback.ai/api/knowledge.json` — full content graph - `https://qback.ai/api/profile.json` — schema.org Person - `https://qback.ai/api/builds.json` — projects - `https://qback.ai/api/newsletter.json` — newsletter archive - `https://qback.ai/.well-known/mcp.json` — MCP discovery manifest