Every surface is a real, working thing. Below: the fifteen that are live, and the three still on the drawing board, each with its source and thinking attached so the work can be picked up again.
Everything here answers 200 and does something. Links open the live product. Where a repo is public-facing it is linked; the rest are private and marked with a lock.
Your own AI agent handles your inbox, briefs your week, and chases unpaid invoices, with a spending ceiling you set. No terminal.
Immersive, alive stories for ages 5 to 7. An 11-agent team writes, narrates, and scores each one.
Try a story →Generate structured study notes from any topic, on an AWS-native pipeline.
Open Study Notes →A companion made of memory and direction, and why it does not need a soul.
Read the paper →Dimensional reasoning and the root of all symbols. A vision essay.
Read the essay →Igbo divination.
Open Agb'áfa →Done-for-you front desk for plumbers, HVAC, electricians, and handymen. Every text answered in under a minute, jobs booked, you get the summary. $497 setup, $197/mo.
See Fronte Reply →A real AI employee for your small business, set up for you and working around the clock. $149 setup, $49/mo.
See the offer →A personal AI helper that remembers you and does the small stuff: email, lists, reminders, forms, letters. $19/mo.
Open Box →Many thoughts, one board. Multi-stream thinking for OpenClaw.
Open OpenStream →Requirements in, verified job plans out. The first dandelion machine.
Open the Planner →The machine factory (Dandelion). A machine is a model plus a curated skill surface plus a gate plus memory, scoped to one job.
Open Machines →Tick as you go. A 2026 to 2030 planning calendar.
Open the Calendar →A proof of concept: three working steps, on real data.
Open the Dashboard →Staging copies live at dev.stories.fronte.io and dev.study.fronte.io. One more surface, Voucha, is live but early — it is a build still in progress, so it sits below with its source.
Three projects are mid-thought. Each dossier below carries what it is, its live surface, the source and key docs on GitHub, and how the thinking evolved. Open a repo and pick the thread back up.
What it is. It began as a voucher and redemption mechanic on top of Boxes. It grew into its own product: Fronte Income, a social platform where a person sells a slice of their future income to backers as a legally binding contract, with the feed ranked by how much skin each voice has in the person it talks about. The current live surface is a small, early page (Voucha).
A voucher and redemption mechanic bolted onto Boxes became its own product, Fronte Income: sell a slice of future income to backers as a binding contract, feed ranked by skin in the game. Payment rails settled crypto-first, fiat second. The access model settled as "open play, earned admission." Then the drafting ran in parallel: a YC application in the founder's voice, a pricing and unit-economics model, two senior-architect tech plans (A: platform, API, React; B: rails, contracts, security) reconciled through one shared interface file, and a micro-staking "Penny Board" layer that keeps cents-as-play and cents-as-return in two clean layers with one hard rule between them.
Regulatory-first throughout. The Howey analysis and general-solicitation rules are the binding constraint, and every undecided legal call is marked for Dozie and counsel rather than guessed.
Status Early. Live page is a basic-auth stub; design and plans are complete; not for public distribution. Testnet and test-money only.
What it is. A managed AI-agent product for non-technical people: own a private agent that does real work, in about five minutes. Ships at boxes.fronte.io. It is live, and the oldest threads of the design are worth resuming.
Three attempts, converging. First, "boxes" (openclaw-boxes) was built like a founding team: DHH on architecture, Jason Fried on product, a charter, a binding build plan, and a deep design corpus. That work produced the decided compute model (shared host, N isolated agent-units), an honest cost floor near $13/mo per agent, and a one-click funnel with billing and auto-provisioning proven live. It also produced a written mistake list.
It restarted as "aiboxes": a serverless control plane (API Gateway, Lambda, DynamoDB, SQS) with agents running on our own host, and that mistake list carried forward explicitly. That converged into "Fronte Agents" (fronte-agents), where a two-role process (marketer against senior architect) produced a practical MVP spec: a flat subscription plus metered AI usage under a user-set hard cap, with a non-technical person succeeding unaided in under five minutes.
Status Live at boxes.fronte.io. Billing, autonomous provisioning, and the one-click funnel are shipped and proven. Live prod Stripe keys and some hardening items remain open; the funnel has run ahead of the product at points.
What it is. The machine factory. A dandelion machine is a model plus a curated skill surface plus a gate plus memory, scoped to one job. The planner was the first machine; the product became the factory that makes machines.
It started as a frame: a dandelion machine is a model plus a curated skill surface plus a gate plus memory, scoped to one job. The planner was the first machine (requirements in, verified job plans out). Hand-building it did not scale, so the work moved to the factory that makes machines: a spec declares a machine as data, and one generic runtime runs any spec.
The gate became the moat. A machine that cannot refuse its own bad output is just an assistant, and the factory's meta-gate refuses a spec that cannot refuse. The proof it generalises: one runtime running two unrelated machines (planner and triage) from two spec files that share nothing.
Status Live at machines.fronte.io (the factory) and planner.fronte.io (the first machine). The factory runs two machines. The name is held at spec level pending trademark clearance.
ship things people can try today.
fronte.io is Dozie Okeke's corner of the internet, a studio where a handful of AI-native products are built in the open, with Ignacio, an AI co-builder who plans, writes, and ships alongside him. The list is kept deliberately whole: live surfaces, and the ones still mid-thought with their reasoning attached. If you build, write, or just like watching ideas become real things, this is the place.