The one-glance read on who they are and how they grow. Each point is verifiable from the receipts above.
An open-source TypeScript/JavaScript framework for building AI agents, workflows, memory, and RAG, built by the team behind Gatsby, with a paid cloud platform layered on top.
Founder-led, launch-driven growth on top of an open-source developer tool.
Over 1M weekly npm downloads as of June 2026, ~309,000 monthly site visits as of 2026-07-26, and $35M total raised across a 2025 seed and an April 2026 Series A.
Founder and CEO Sam Bhagwat framed the company's spending itself as the pitch: "Mastra doesn't have an office. Instead, we're spending $1.5 million of our seed funding printing and distributing our agents books."
Mastra treats major version milestones, not ongoing campaigns, as its growth lever, relaunching on Hacker News for its original 2025 open-source debut and again for its 1.0 stable release, both times riding Sam Bhagwat's own account to convert a single launch day into a lasting npm-download curve.
The order the channels came online. Sequence is strategy: what they did first, and what they layered on once demand existed.
Estimated demand, the channel split behind it, and the keywords and referrers doing the work. Directional modeling, not audited analytics.
| Keyword | Volume | Weight | CPC |
|---|---|---|---|
| mastra | 31k | $2.01 | |
| mastra ai | 9.3k | $3.18 | |
| opencode go | 290k | $2.00 | |
| opencode zen | 66k | $2.31 | |
| venice ai | 373k | $1.04 |
Mastra pulls ~309,000 estimated monthly visits, down 4.6% over the last 3 months as of 2026-07-26. Search Organic (45%) and Direct (40.2%) together account for 85% of traffic, meaning most visitors either already know the name or find it through search rather than social or paid placement; Referrals trail at 5.5%, with Social Organic, generative-AI referrals, display ads, and email combining for under 10% of the rest. The branded terms "mastra" (~31,300 searches/month) and "mastra ai" (~9,300 searches/month) show the name itself carries real standalone search demand, while the list also surfaces alongside far larger adjacent-tool queries like "venice ai" (~373,500 searches/month), a rival AI platform's own name, underscoring how much bigger the surrounding demand pool is than Mastra's branded slice of it. Referring traffic includes ycombinator.com, consistent with its YC W25 batch placement, alongside sourcescrub.com and force.com, which read as B2B data-enrichment and sales-prospecting crawlers rather than genuine reader referrals. Its closest listed competitors mix direct rivals like opencode.ai with ai-sdk.dev, the Vercel toolkit Mastra was originally built on top of per its own launch post, blurring the line between competitor and dependency, plus inngest.com, a durable-workflow-execution player adjacent to Mastra's own workflow engine.
The specific pages earning their organic search traffic, and the pattern behind why they rank. Adapt the format, not the topic.
The clearest organic signal is the homepage itself, which ranks #1 for "mastra" and is the dominant organic entry point into the site. Beyond that, Mastra is seeding a programmatic "models/providers" directory, individual pages for LLM provider integrations like nano-gpt and iflowcn, that target long-tail, provider-name searches ("iflow" and "nanogpt" both rank around position #10). It's an early-stage version of the classic programmatic-SEO pattern for developer tools: one page per integration, compounding as more providers get added, though these specific pages are drawing minimal traffic so far.
Not traffic share. How much weight the growth system actually puts on each channel, with a one-line read on the role it plays.
Open roles read like a roadmap: the functions they’re staffing show where the company is investing next and what stage it’s at.
For founder-led SaaS the breakdown shifts from ads to traction: where the first users came from, how the founder grows it in the open, and the compounding organic surface.
Mastra's user base grew almost entirely off two founder-run Hacker News launches a year apart, with a viral regional spike in between that proved the first spike wasn't a fluke.
The GitHub spike: In the week after the original Show HN debut, Mastra's GitHub stars jumped from roughly 1,500 to about 7,500, converting a single front-page day into a lasting repo-adoption curve rather than a one-day traffic blip. The Japan inflection: In March 2025, CEO Sam Bhagwat posted that Mastra "has gone viral in Japan," with 16% of web traffic suddenly arriving from the country, an unplanned regional pickup that showed the framework was spreading through developer word of mouth well beyond the original English-language HN audience. The relaunch: Mastra treated its stable 1.0 release as a second launch rather than a routine changelog entry, going back to Hacker News with a post that explicitly called back to the original submission from a year earlier, a continuity move that reset attention on the same platform that built the initial audience. Early npm adoption climbed into the multi-hundred-thousand weekly range by the time of the 1.0 relaunch, the download curve that later crossed 1 million weekly downloads.
Build in public at Mastra runs through three personal X accounts rather than one brand voice, with CEO Sam Bhagwat carrying the largest reach.
Three founders, three feeds: CEO Sam Bhagwat (15,684 followers, ~14,676 posts), CTO Abhi Aiyer (2,507 followers, ~3,994 posts) and CPO Shane Thomas (4,426 followers, ~5,574 posts) each post from personal, bio-confirmed accounts tied to the company rather than funneling updates through a single company handle. Customer-win posts, not just metrics: Shane Thomas shared a customer's own words on switching from LangChain, "all we wanted in ai was in mastra," a social-proof post distinct from the milestone announcements Sam typically runs. Milestones over day-to-day updates: the pattern skews toward funding and download-count posts rather than routine behind-the-scenes updates, reinforcing the read that Mastra's public-building cadence is event-driven. No founder presence on Reddit: the founders don't post there themselves, so Reddit activity around Mastra stays limited to the third-party comparison and build-log threads already noted, leaving X as the only channel where the founders directly shape the narrative.
Organic search sits at 45% of Mastra's traffic mix, but the underlying content system is still a narrow, early-stage bet rather than a mature compounding engine.
Docs as a product surface: Mastra shipped its documentation as an installable MCP server that IDEs like Cursor and Windsurf can query directly, turning static docs into a channel embedded inside the coding tools developers already have open, rather than a page they have to go find. Programmatic pages, still thin: the provider-integration pages are a templated, one-page-per-integration structure that mechanically compounds as more LLM providers get added, an early instance of the classic programmatic-SEO pattern for dev tools rather than a fully built-out library yet. No paid acquisition: there's no visible Meta, Google, or TikTok ad presence in the evidence, meaning the entire funnel runs on launches, founder posts, and the organic-search share above, a deliberate-looking bet on owned and earned surfaces over paid demand generation.
The channels are not separate. They are one system where each stage feeds the next. Here is the read, then the plays to run tomorrow.
Mastra's loop starts with a founder-driven launch spike, sustains itself through personal-account visibility between launches, and converts a slice of free open-source users into paid cloud accounts.
the free, self-hosted Starter tier ($0/month, pay-as-you-go credits) is the top of funnel for every npm install, while production teams convert into the $250/month Team tier or custom Enterprise deals, the same layer that Marsh McLennan (an internal search tool used by 75,000+ employees) and Brex (agents used in its $5.1B Capital One acquisition process) are cited as running on. Founder reach as the connective tissue: between launch spikes, the three personal X accounts above are what keeps the download curve visible rather than any always-on brand campaign. A trust gap: the June 2026 npm supply-chain attack generated real inbound attention from security press and creators, but there's no visible crisis-communication content in the evidence addressing it directly, leaving a trust-repair gap open after a genuinely reputation-shaking event.
free-to-paid conversion from the Starter tier into the $250/month Team plan, actual credit/CPU-hour consumption and overage revenue per account, and contract size for the named enterprise accounts like Marsh McLennan and Brex.
The proofMastra's two Show HN launches a year apart each reset front-page attention, with the second explicitly calling back to the first.
The adaptationTreat your next major version, license change, or benchmark win as its own launch, not a changelog line. Write a new launch post that opens with a callback to your original one ("a year ago we shared X here, here's what changed"), and publish it the same day you ship the update. Start by drafting that callback line now, before the next version ships, so it's ready the day you need it.
Cost: $0 · Time to signal: days · Works pre-PMF: yes
The proofSam Bhagwat's personal account, not a Mastra brand handle, carried the funding and download-milestone posts that drove the widest reach.
The adaptationWhen you hit a real number worth sharing, a user count, a funding close, a download milestone, post it from the founder's personal account in a plain "we crossed X" format rather than routing it through a company page. First step: write the template now ("we just crossed [metric]") so it's ready to fire the moment a real number lands.
Cost: $0 · Time to signal: days to weeks · Works pre-PMF: conditional, only once there is a genuine number worth announcing
The proofMastra built its documentation into an MCP server so coding assistants like Cursor and Windsurf can query it directly, rather than leaving it as a page developers have to search for.
The adaptationWrap your existing docs or API reference into a minimal MCP server (or equivalent plugin) exposing a single "search docs" tool, publish it as its own package, and link it in your README. Start by picking your five most-viewed doc pages and exposing just those through the tool first.
Cost: under $500 · Time to signal: weeks · Works pre-PMF: yes
The proofMastra's provider-integration pages are a single template applied once per new LLM provider, compounding the site's indexable footprint as providers are added.
The adaptationBuild one templated landing page per integration or adjacent tool your own users already search for, reusing one layout so a new page takes minutes, not a new project. Start by listing the ten tools your users mention most often and generating the first three pages from a shared template this month.
Cost: $0 to under $500 · Time to signal: months · Works pre-PMF: yes
The proofA single founder post noting 16% of Mastra's traffic had shifted to Japan turned an analytics anomaly into a public growth story rather than a private data point.
The adaptationCheck your referrer and geography data on a regular cadence, and if any single country or channel crosses roughly 10% of traffic unexpectedly, treat it as a lead: translate your top landing page into that language and post about the spike itself. Start by pulling your current geographic breakdown and flagging anything already near that threshold.
Cost: under $500 · Time to signal: weeks · Works pre-PMF: yes Not transferable at an earlier stage: relaunching a Show HN post only works if the first one already built real credibility, and Mastra's founder-led reach on X was built on a first launch that had already gone viral, neither is available to a founder without a launch moment behind them yet.
Every morning we take one company that is actually growing and break down where its customers come from: the ads still running after a year, the channel doing the real work, and the play you can run this week.
Systemaic · directional intelligence. Traffic, spend, and reach figures are SimilarWeb-style estimates and qualitative reads of public data, not audited numbers. Built on real public receipts.
Launch archives: Hacker News: Multiple mastra NPM packages compromised · Hacker News: Show HN: Mastra – Open-source JS agent framework, by the dev · Hacker News: Show HN: A “Course” as an MCP Server · Hacker News: The MCP Registry Registry · Hacker News: Building AI Agents with Mastra and Deno · Hacker News: Mastra is now Apache 2.0 licensed
Researched facts: Company tagline / positioning: The Javascript framework for building AI agents, from the G · Legal entity name: Kepler Software, Inc. · Headquarters address: 2261 Market Street Ste 5559, San Francisco, CA 94114 · Public launch (Show HN): Launched publicly on Hacker News around Feb 22, 2025, as an open- · Seed round: $13M seed round from 120+ investors including Y Combinator, Paul Graham, Gradi · Series A round: $22M Series A led by Spark Capital, bringing total funding raised to $35M · Series A blog post headline: Mastra positioned the raise as being 'to help every developer · 1.0 stable release: Mastra reached its 1.0 stable release on Jan 20, 2026, locking APIs an · Pricing — Teams tier: Teams Platform plan costs $250/month, including SSO and SOC 2 docs,
Community threads: Y Combinator Launch / Hacker News: Founders positioned Mastra explicitly as the anti-LangC · Company blog (Convex) / discussed on Hacker News: A Convex engineer spent a week reimpleme · Hacker News: An HN thread questioning Mastra's headline LongMemEval benchmark score reflec · Security press / Hacker News: North Korean-linked actors compromised the @mastra npm scope · DEV Community: A developer's direct before/after comparison (41 hours fighting LangGraph's · Hacker News: A grassroots HN 'discovery' thread shows organic surprise at how under-the-ra · GitHub: A neutral third-party benchmark repo pitting the two frameworks head-to-head found · Case study (CodeRabbit): Mastra's own engineers admitted, in a vendor case study, that AI-
Traffic, spend, and revenue figures are estimates as noted in the report; the links above are the primary public artifacts.