Dev Tool GTM, Done Agentically
  • Home
  • AI
  • Dev Tool GTM, Done Agentically
By Tessa Kriesel profile image Tessa Kriesel
9 min read

Dev Tool GTM, Done Agentically

My entire marketing operation at Tabstack is agentic, and I built it like an engineer. One source of truth in Quiver. Nothing hard-coded. Every context change versioned and approved. Agents do the work and I make the calls. Here's how it runs.

I'm Founding Technical GTM at Tabstack. I run GTM with a stack of agents and tools, and I stay in the loop for the calls that need a human.

This post is the whole system: the parts of dev tool GTM I run, how I do each one today, and what comes out of it.

One rule holds it together. Positioning, ICP, messaging, voice, and vocabulary are never hard-coded anywhere in the stack. Every system reads them from Quiver context at the moment it works. Change the context in Quiver once, and Tabstack OS, Hermes, and every skill follow on their next run.

The systems

  • Quiver: the system of record. It holds the product marketing context (positioning, ICP, messaging, competitive landscape, customer language, brand voice, words to use and avoid), plus research, campaigns, the content workflow and calendar, and every report the agents produce. It's versioned. Agents propose changes and I approve them.
  • Tabstack OS: where I work. A Claude Code workspace that loads Quiver context every session, drafts and reviews content and docs, queries analytics and the CRM, and runs the other systems.
  • Hermes: my always-on agent. It runs scheduled jobs for competitive intel, community listening, prospecting, sponsorships, social drafts, and event briefs. It reads Quiver context before every job and saves every report to Quiver. It drafts and never posts, publishes, emails, or contacts anyone.
  • Rival: my open-source competitive intel app. I built it on Tabstack to dogfood and learn the product.
  • Magister: weekly AI visibility tracking and AEO audits.
  • MentionDrop: social listening on the forums and social sites our own tools can't reach reliably.
  • PostHog and GA4 via Supermetrics: product and web analytics: signups, API usage, activation, and retention.
  • Airtable CRM: data-collection-approved signups have an account, with lifecycle stage, usage, acquisition source, and a PostHog event timeline.
  • Granola and Linear: call transcripts, and where product feedback lands for engineering.

At a glance

Marketing effort How we do it today Result
Customer and market research Every signal, call, and thread lands in Quiver research, which extracts themes, sentiment, and quotes, scores it against our hypotheses, and proposes context updates. Customer language and hypothesis evidence in one place. Approved proposals update what every agent reads.
Positioning Human-led. I make the call from Quiver research, product data, and customer calls. One positioning that Tabstack OS, Hermes, and every skill read from Quiver at runtime. Change it once, everything follows.
Messaging Agents draft from Quiver context. I polish. One vocabulary across site, docs, and posts, with the banned-word list enforced on every draft.
Competitive intel Rival, my open-source app built on Tabstack, scans 15 competitors across 11 page types (homepage, pricing, changelog, docs, careers and hiring tech stack, blog, G2 and other reviews, social, GitHub, company profile, tech stack), keeps every diff, and writes intelligence briefs with threat tiers and 7-axis scores. Hermes reads it with Quiver's watchlist every morning, drops the noise, and scores each change against the positioning in Quiver context. A daily brief of what actually changed and what to do about it. This week: check Parallel's new Task API tiers before writing Research comparison copy.
Community listening Find the conversations where the ICP in Quiver context hits the problems we solve, so we show up where the pain is and learn their language. MentionDrop covers the sites our own tools can't reach reliably. Hermes turns the daily haul into exact threads with draft replies in my voice. 2,680 mentions a month filtered to a short daily list. I reply the same day; the best threads become research.
Content production workflow Tabstack OS drafts, runs the editorial principles review, fact-checks against the live API, and renders covers. Quiver runs draft, review, approve, publish, plus the calendar. 13 posts in the 24 days after the pivot, every one technically verified.
Docs and API reference Tabstack OS tech-writing, api-reviewer and docs-reviewer agents, sdk-sync to track SDK changes. Docs keep pace with every SDK release.
Social Hermes drafts LinkedIn and X posts every Thursday from the Quiver calendar. Native drafts waiting for every published post. I edit and post.
Prospecting and CRM Hermes prospects every Tuesday into Airtable. crm-import and trace enrich signups. My network: 10k people I've connected with over nearly 20 years in tech in one deduplicated database, which supplies prospects, warm intros, and contact emails. 5 to 10 qualified teams a week, each with a cited signal, deduped. 491 signups enriched. Contacts for outreach in minutes.
Sponsorships and creators Hermes finds creators and newsletters whose audience matches the ICP in Quiver context. A ranked sponsorship shortlist every week.
Events and CFPs Hermes writes a Monday brief, scored attend, speak, or sponsor. Events 30 days out and CFPs 90 days out, already scored.
Website Copy and site architecture from Tabstack OS, Astro build, launch audit. Rebrand live 21 days after the positioning decision was. made.
Analytics and attribution PostHog and GA4 (via Supermetrics), queried in plain language from Tabstack OS. Answers in minutes, no dashboard requests.
Activation and retention PostHog tracks every account from signup to API key to API usage, plus retention at organization level. The Airtable CRM carries each account's milestone timeline, acquisition source, and use case. This data, with Quiver research, is why we pivoted.
AI visibility (LLM citations) Magister tracks 15 prompts weekly across Claude, ChatGPT, Gemini, and Google AI Overviews, plus AEO audits. Visibility score 22.3 to 42.7 in two weeks. Homepage AEO 80 to 95.
Product feedback loop Fully automated. When any of my agents hits a Tabstack error or DX problem, in any repo, it logs it, and a hook files it as a Linear ticket to engineering. No human in the loop. Every bit of friction my agents hit becomes an engineering ticket, without anyone stopping to write it up.

The eight cores of dev tool GTM, and how the stack covers them

1. Know the developer

  • ICP and jobs to be done from real interviews: Granola captures calls. Each call becomes a Quiver research entry, which is processed automatically for themes, sentiment, quotes, and whether it validates or challenges an active hypothesis, then turned into proposed context updates. I approve them.
  • Events and in-person community to validate the market: conversations at events and meetups go into Quiver research from my own notes, then get the same automatic processing: themes, quotes, hypothesis signals, and proposed context updates.
  • Watch developers try the product and log the friction: my agents use Tabstack every day. When one hits an error or DX problem, it logs it and a hook files it as a Linear ticket to engineering. Fully automated.
  • Track competitors and real alternatives, including build-it-yourself: Rival scans 15 competitors across 11 page types and writes threat-scored intelligence briefs. Hermes reads it with Quiver's watchlist every morning and scores each change against the positioning in Quiver context. Community listening surfaces the DIY builds people post.

2. Positioning and messaging

  • Category, alternatives, differentiated value: human-led. I make the call from Quiver research, product data, and customer calls. It lives in Quiver context.
  • Messaging hierarchy: messaging pillars live in Quiver context. Agents draft from them. I polish.
  • Homepage, docs landing, and pricing copy: Tabstack OS drafts from Quiver context and writes the site architecture. The rebrand went live 21 days after the positioning call.
  • Everyone saying the same thing: every agent and skill reads Quiver context at runtime and enforces its words-to-avoid list on every draft. Context changes are proposals I approve, so nothing drifts silently.

3. Developer experience

  • Time to first value: PostHog tracks every account from signup to API key to API usage. The Airtable CRM carries that milestone timeline per account, with usage days and span, so I can see how far each signup got and how long it took.
  • Docs, examples, SDKs, error messages: Tabstack OS writes docs, READMEs, changelogs, and API reference. Review agents check every claim against SDK source and the live API. sdk-sync tracks upstream SDK, API, and docs repos and updates what changed.
  • Agent-readiness: llms.txt and llms-full.txt on the site, plus AEO audits in Magister.
  • Route field feedback into the roadmap: the automated DX-notes-to-Linear loop, plus product signals flagged on Quiver research entries.
  • Pricing and packaging: competitor pricing changes arrive daily from Hermes. Pricing decisions are fed from research, made by a human.

4. Awareness and demand

  • Technical content: Tabstack OS drafts, reviews against the 10 editorial principles, fact-checks against the live API, and renders covers. Quiver runs draft, review, approve, publish, and the calendar. 13 posts shipped in the 24 days after the pivot.
  • Search visibility, including AI answer engines: Magister tracks AI visibility weekly across Claude, ChatGPT, Gemini, and Google AI Overviews, plus AEO audits. Search Console through Supermetrics.
  • Community presence: MentionDrop listens on the forums and social sites our own tools can't reach reliably, using keywords tuned to the ICP in Quiver context. Hermes turns the daily haul into exact threads with draft replies in my voice. I reply. The best threads become research.
  • Events, talks, CFPs: Hermes writes a Monday brief of events 30 days out and CFPs 90 days out, scored attend, speak, or sponsor.
  • Sponsorships: Hermes finds creators and newsletters whose audience matches the ICP in Quiver context.
  • Integrations, partnerships, listings: tracked as Quiver campaigns. Directory submissions and co-marketing skills in Tabstack OS.
  • Social: Hermes drafts LinkedIn and X posts every Thursday from the Quiver calendar. I edit and post.

5. Activation and conversion

  • Instrumentation: PostHog and GA4, queried in plain language from Tabstack OS, which also builds the dashboards.
  • Finding and fixing the biggest drop-off: the funnel from visit to signup to API key to usage is read in PostHog. Every account carries its acquisition source and use case, so drop-off splits by channel and by audience.
  • Retention: PostHog tracks retention at organization level. The CRM shows usage days and span per account.
  • Product-qualified accounts: the Airtable CRM holds every signup with lifecycle stage, usage, acquisition source, newsletter status, and a PostHog event timeline, kept current by crm-import. Accounts that created an API key are enriched through trace and classified by use case. Hermes adds qualified prospects weekly, each with a cited signal, deduplicated.
  • Network intelligence: my network: 10k people I've connected with over nearly 20 years in tech into one deduplicated database with append-only enrichment (Hunter, GitHub, Gravatar, Tabstack). Searchable in a local UI. It's where prospects, warm intros, and contact emails come from. Hermes keeps it up-to-date.
  • Lifecycle email: drafted in Tabstack OS and tracked as a Quiver campaign so performance can be logged.

6. Revenue

  • Sales enablement: one-pagers, objection handling, and demo material drafted in Tabstack OS from Quiver context.
  • Founder-led sales: Quiver holds the ICP shape. Hermes finds teams that fit it in my network and beyond, and adds them to the CRM with the signal. I reach out.
  • Customer stories (in progress): drawn from Quiver research, the same calls, quotes, and notes that feed everything else.

7. Measurement and ops

  • Decisions from the data: our most recent pivot came out of this system. PostHog activation and retention data, plus Quiver research from calls, interviews, and forum threads scored against active hypotheses, is why we repositioned.
  • Data that keeps getting better: I've improved the data the whole time. Client-side web analytics added in June. Accounts tagged by job (extract, research, automate) from endpoint usage, first-touch source backfilled, and PostHog events synced into the CRM. Analytics identity and signup-source tracking shipped for the pivot. Complete attribution tracking.
  • Weekly review: every Tabstack OS session opens with a brief pulled from Quiver: what's due, Hermes reports and research waiting on review, what's publishing, what shipped, and pending context proposals.
  • Experiments with stated hypotheses: active hypotheses live in Quiver context. Every research entry is scored against them.
  • Metrics: a north-star metric and the funnel behind it in PostHog, plus GA4, Search Console, and Magister AI visibility, all queried from Tabstack OS and synthesized into Quiver as performance logs and the weekly marketing report.
  • Tooling, budget, hiring: drafted in Tabstack OS. Decisions are human.

8. Internal alignment

  • One agreed ICP and priority set: Quiver context is the single source, versioned, with every change approved by me. Everyone works together in Quiver.
  • Report in business terms: exec updates, one-pagers, and decks drafted in Tabstack OS.
  • Weekly marketing updates: built and sent on Mondays by Hermes.

What stays human

Agents draft. People decide. Positioning calls, design, privacy, publishing, posting, and outreach all stay with a human.

If this stack is speaking your love language, check out Quiver—#3 on Product Hunt on Vercel Day. 🔥

You might also like How I Grew Signups 71% in 90 Days by Running GTM like an Engineer.

By Tessa Kriesel profile image Tessa Kriesel
Updated on
AI Developer GTM DevRel & Community Coding & Dev

Read Next