0809 | Basedash Subscriptions, Toolport, Hexis, GTM Co-Founder

Mia: Welcome back to ProductHunt Daily, I'm Mia.

Milo: And I'm Milo. We're Bri's daily look at the launches shaping how teams build, ship, and sell — here's what's moving today.

Mia: First up, Basedash turned its business intelligence platform into a subscription business with Basedash Subscriptions.

Milo: Then Toolport, a free open-source gateway that gives all your AI agents one local connection point.

Mia: And from Bevel, Hexis is open-sourcing the core of its platform — Git-backed skills for AI agents.

Milo: Plus, Shane O'Connor's open-source go-to-market skills for solo teams, and a closer look at AstraPixels, today's single-product story.

Mia: That's this episode of ProductHunt Daily. Let's dig in.

Mia: Basedash, the AI-native business intelligence platform, has just shipped a feature called Subscriptions, and it targets a very familiar pain. Founder and CEO Max Musing says the reporting ritual — the Monday metrics review, the Friday pipeline check, the first-of-month board pack — always needs someone to remember to open the dashboard, take a screenshot, and post it before the meeting. Subscriptions remove that errand entirely, so the report comes to you instead of you going to it.

Milo: So how does it actually work? You subscribe to any dashboard or chart, pick a schedule, and get a fresh snapshot delivered on that schedule — straight to an inbox or a Slack channel. And here's a nice touch: scheduling is written as a plain sentence, like every week on Monday at 9 AM. You can pick every day, weekdays, a specific weekly day, or monthly and quarterly on a date, at whatever hour you like. So schedules run from daily all the way up to quarterly.

Mia: And the delivery options are genuinely flexible. Email can go to a set of teammates, or the report can land in a Slack channel, or both. In Slack, the report posts right to the channel with each chart embedded as an image — no clicking through, no logging in. In email, recipients get charts inline plus an Open dashboard in Basedash button that links to the live version.

Milo: The key safeguard, though, is that charts are rendered at delivery time from live data, so a subscription never sends a stale report. It's always current when it arrives. And the feature is designed to stack on a single dashboard. Think of one revenue dashboard getting a daily digest to exec team inboxes, a Monday morning post in a metrics channel, and a monthly board snapshot — three subscriptions, three schedules, three audiences all built on the same dashboard.

Mia: That means the same underlying report can serve completely different cadences and audiences without anyone having to maintain three separate versions. Each subscription is managed from the dashboard or chart it belongs to, so everything stays in one place. For anyone who's ever been the de facto report-maker on a team, this is exactly the errand that quietly eats up time every week.

Mia: Let's move to a different kind of tool — a local gateway called Toolport that tackles a messy part of working with AI agents. It's free, open-source, MIT-licensed, and fully local, running on Windows, macOS, and Linux with no account and no cloud involved. The pitch: set up each MCP server once, and every AI agent shares it — instead of pasting the same API keys into separate config files for Claude, Cursor, VS Code, Codex, and others. Client support runs up to 33 apps.

Milo: Why does that matter? Creator Tyler says every client wanted its own JSON file with the same API keys stored in plain text, and every MCP server dumps its full tool list into context on every request. He says three servers cost him about 24,000 tokens of definitions before he'd even asked anything. So Toolport sits in between as the single gateway, holding the keys once and letting every agent share them.

Mia: The real differentiator is something called lazy tool discovery. Instead of exposing every tool definition up front, Toolport exposes just a handful of meta-tools that the agent searches on demand. The product site shows tool-definition tokens on one request dropping from roughly 23,700 without Toolport down to about 886 with it.

Milo: And the maker reports some striking benchmark numbers off the back of that — though it's worth flagging these are self-reported, with no independent verification. Claimed figures include up to 91 percent fewer total tokens at the same task success, roughly 74 to 91 percent fewer tokens across servers, and 99.5 percent lower definition overhead on a real catalog of 415 tools.

Mia: Given that keys are the sensitive part here, security was a core design choice. API keys live in the OS keychain, get injected at runtime, and never sit in config files or the cloud. On top of that, Toolport fingerprints every tool and flags any that look like a rug pull — a safety check on top of the efficiency win.

Milo: So the practical takeaway is clear. For anyone juggling multiple AI coding clients, Toolport removes the duplicated key mgmt and trims what's essentially wasted token spend on tool definitions. Those efficiency numbers are the maker's own claim, but the direction — fewer tokens on every request — is the obvious near-term benefit for anyone running several agents day to day.

Mia: Let's start with Hexis, an open-source project that just launched on Product Hunt. It's built by a company called Bevel, and it's essentially positioning itself as the Git-powered layer for AI agents — the part that holds the skills, tools, and context those agents use. The maker describes a real pain point behind it: teams, including their own, keep throwing agent skills and context into GitHub, where versioning and pull requests work fine for a while. But then the problems pile up. Non-technical people face a steep learning curve, file-based access management doesn't really work on GitHub or GitLab, and simply sharing a skill with a coworker turns into something slow and clunky.

Milo: So what does Hexis actually change about that workflow?

Mia: It keeps the parts of Git that work — versioning and pull requests — but layers a friendlier, non-developer interface on top. So instead of wrestling with Git, an admin decides which tools, skills, and context each person, team, or agent can access. Users then connect whatever agent they want — Claude Code, ChatGPT, Cursor, any agent that speaks the open MCP protocol — without ever touching a Git command. They can suggest changes or submit new skills directly. And under the hood, all of it lives in a git repository the company owns, on any host you like, from GitHub and GitLab to Bitbucket, Azure DevOps, or self-hosted. Skills live as folders, tool manuals come with an encrypted secrets vault, and there's support for branches, change requests that need owner approval, role-based access, and a built-in remote server using OAuth 2.1.

Milo: That addressable access control piece is interesting, because rules apply to agents exactly the way they apply to people. So an agent doesn't get to see more than the team member it's acting for would.

Mia: Exactly. And because it all runs on open protocols, there's a real vendor-agnostic angle here. Teams can switch agent vendors based on price and performance — or mix agents by task and role — without rebuilding everything their agents already know. The knowledge, skills, and context stay in the repo instead of being locked into one assistant. Employees connect once, in a matter of minutes, and that workspace follows them across whichever agent they're using at the time. For teams that have been stashing their agent setup in a shared repo, the pitch is essentially: keep your versioning and ownership, but get a proper interface and real access control on top.

Mia: The second launch is The GTM Co-Founder, built by Shane O'Connor. It's an open-source, MIT-licensed set of go-to-market skills aimed squarely at solo technical founders building dev tools. And the maker's framing is that it's different from the generic sales and marketing advice most AI assistants dispense. Instead of broad platitudes, it interviews you once, learns what your actual product and market look like, and then produces a prioritized go-to-market roadmap — who it's for, positioning, your first fifty users, launch, and pricing.

Milo: And it's explicitly grounded in real-world playbooks rather than generic advice?

Mia: Yes — the maker says it draws on the playbooks of Adam Frankl and Jakub Czakon, plus their own time as a founding account executive and go-to-market advisor. It's free, it's open source, and installation is a single command that adds it as a skill, requiring Node.js. The positioning line is memorable, too: the maker says your agent is too nice, and this is the go-to-market co-founder that says what others won't.

Milo: Have people actually tried it and shared reactions?

Mia: There's at least one community commenter who ran it on a beta product and reported excellent and detailed analysis, saying they planned to use it often. The other comments mostly raised open questions rather than confirmed flaws — which is worth noting. One is how the tool keeps context consistent across positioning, pricing, and launch, so its recommendations don't start contradicting each other over time. Another is how it handles knowledge drift as the product and market change. And there's a bigger category question: whether skill-as-a-product is a growing category at all, and whether this tool's reach is tied to how widely people adopt Claude Skills specifically. So the constructive reactions are warm, but the open questions are about consistency over time and whether the whole distribution model is sustainable.

Milo: On the roadmap itself, one commenter asked whether it's mostly pre-launch positioning or whether it also tackles the gap between nobody linking to generic content and content that actually earns attention. That feels like the more valuable question for a solo founder — is this a launch checklist, or does it keep working after day one?

Mia: Right — and that's where the state of the discussion stands. The tool is less than a week old in this conversation, still early, and the practical consensus so far is that it produces genuinely useful analysis and prioritized output. The open questions are about longevity: whether its context stays coherent across decisions, and whether it grows with your product after that initial interview. Either way, it's free and open source, so for a solo dev-tool founder the cost of trying it is just the time to run the interview and see if the roadmap matches how you actually think about your market.

Mia: There's a new launch on Product Hunt called AstraPixels, and it's a bit unusual. It's an animated, pixel-art map of the solar system, built by a maker who goes by @nicklaunches, live at astrapixels.com. The core pitch is that it places every body at its real current position, computed from published orbital elements with an astronomy engine, not copied from a pre-made list.

Milo: That's the part that makes this interesting. You can zoom from the Sun all the way out to the Kuiper belt, browse a catalog of 171 objects each with sourced facts attached, and even wind the clock forward a week to see where things will be. There's also a sky-events page that works out the next six months of conjunctions, oppositions, meteor showers, and eclipses.

Mia: And the maker is pretty explicit that this is meant to be an astronomy tool first, not just a gimmick. There's even a "true scale" toggle that shows you how much the map is lying to you about distances. So who's it for? People who want something playful but still instrument-like, to really see where the planets and asteroids actually are right now.

Milo: But here's the famously clear caveat. On the business side, the maker says they built the whole commercial setup around the three ways a Million Dollar Homepage clone normally dies. First, inventory. The main asteroid belt is about three thousand procedurally generated rocks, so the five-dollar asteroid tier can't sell out. Genuinely scarce items, like the Sun, the eight planet skins, and the comets, are leased rather than sold. A moon runs forty-nine dollars.

Mia: So if you claim an asteroid, it costs five dollars once, you keep it forever, and you can name it whatever you like. You can even advertise on a claimed object. And second, there's the retention angle, which loops right back to the tool part. The map, the sky events, the time-winding, the object catalog, all built to keep people coming back rather than just buying once and leaving.

Milo: And the third way these projects usually die is sign-up friction, which the maker pretty aggressively dodges. No account registration at all. In the maker's own words, Stripe asking for your email is the whole account system. That's it.

Mia: One note for anyone on the fence: the supplied evidence here is a single-product look, so the buyer review caveat is unusually clear-cut. It's a fun, instrument-flavored way to check where planets and asteroids actually are, with a believable inventory approach behind it, but it comes with that honest, the-maker-told-you-the-business-risks framing attached.

Mia: So today we covered the AI launch scene across business intelligence, developer tools, and go-to-market plays. Basedash rolled out a new subscriptions offering to make AI-native BI a real product people pay for.

Milo: Exactly. And on the developer side, Toolport gives every AI agent one local gateway, free and fully open source. Plus Hexis brings a Git-backed skills library to your agent tools, and the GTM Co-Founder hands solo founders a ready-made set of sales and marketing skills.

Mia: And we closed on AstraPixels, our single deep dive for this briefing. That's it for this one — thanks for listening, and we'll catch you next time.