One brain. Every product.
Hivemind OS is the operating system that powers every Taskclan product — shared identity, billing, events, and a hive of specialized AI agents — so each new product launches as an app on a platform, not a company built from scratch.
Built once. Shared by all.
What’s running today
| 338 | 21 | 8 | 5 | 2,400+ |
| Intents | Named agents | Workflows | Products on the runtime | Users on the flagship |
All numbers live from production. Verify any of them on the Live links below.
What is Hivemind OS?
Every AI product needs the same backend before it ships anything that matters: auth, billing, agents, memory, model routing, workflows, knowledge, deployment, usage metering.
We built it once.
Now five of our own AI products run on it — and we’re opening it as a platform for anyone building AI-native software.
One intent. One API. Everything.
Every backend capability is one request:
POST /api/hive/v1/intentNo new endpoints per feature. No new SDK methods to learn. Every product — ours or yours — talks to the brain through the same typed call.
The stack
┌─────────────────────────────────────────┐
Products │ GameNova · Nani · VendorFlyer · … │ ← your app here
└────────────────────┬────────────────────┘
│
▼
┌─────────────────────────────────────────┐
Hivemind │ Agents · Skills · Workflows · Memory │ ← the hive
│ Insights · Tools · Marketplace │
└────────────────────┬────────────────────┘
│
▼
┌─────────────────────────────────────────┐
Platform │ Identity · Billing · Events · Wallet │ ← shared services
│ AI Gateway · Notifications · SDKs │
└────────────────────┬────────────────────┘
│
▼
┌─────────────────────────────────────────┐
Infrastructure │ Supabase · Heroku · R2 · OpenRouter │
└─────────────────────────────────────────┘The hive
Hivemind isn’t a router that calls models — it’s an organism of specialized faculties (we call them minds) that route, plan, decide, generate, coordinate, remember, and speak. Every intent is tagged with one or more minds; the platform schedules, observes, and learns across them.
Seven minds
perception · planning · decisioning · generation · coordination · memory · voice
What you get
- 21 named agents — each a persona-scoped grouping of intents (run with
runAgent(name, input)) - 338 intents — every capability behind a single typed API
- A tool registry — connect external services (Gmail, Slack, GitHub) once, reuse everywhere
- A prompt registry — versioned prompts, A/B variants, eval hooks
- Shared memory — read/write across products with privacy invariants
- A cross-product learning loop — one product’s aggregate insights retune another’s defaults before the agent even acts
Apps on the OS
Five products running on the same runtime, today:
| Product | What it is | Live URL |
|---|---|---|
| GameNova | AI game creator (2,374 users · 3,774 games created) | gamenova.io |
| Nani | Voice-first family assistant | iOS App Store |
| VendorFlyer | AI promo-kit generator (NGN) | vendorflyerai.com |
| NaijaCV | AI career-kit generator (NGN) | naijacv.ai |
| Proposal AI | AI proposal generator (NGN) | proposalai.com |
Each one was easier to build than the last — because the platform handled auth, billing, agents, memory, and model routing. The fifth product cost a fraction of the first.
Hivemind Cloud — the control plane
One dashboard. One login. One bill.
After install, your users sign into Hivemind Cloud to manage orgs, install agents from the marketplace, see usage + billing, and invite teammates — all across every Hivemind product they use.
| Surface | What it does |
|---|---|
| Overview | Org activity at-a-glance |
| Projects | Per-product workspaces |
| Marketplace | 35 first-party listings · install with one click |
| Team | Org members + roles, invite by email |
| Billing | Plan + usage + upgrade flow |
| Developers | API keys, SDK install, registry discovery |
Multi-tenant org model · JIT personal workspace on first login · server-side plan/quota enforcement at the intent layer.
→ Read the Cloud guide · → Open Cloud (live)
For builders
You don’t need a token to install. The packages are public:
# add to your project's .npmrc:
echo "@taskclan:registry=https://npm.pkg.github.com" >> .npmrc
# then install + scaffold:
npm install @taskclan/platform @taskclan/hivemind @taskclan/hivemind-react
npm create @taskclan/hivemind-app@latest my-appThe four packages
| Package | What it is |
|---|---|
@taskclan/create-hivemind-app | Scaffolder CLI — 4 templates (node-starter, react-chat, next-ai-chat, agent-worker) |
@taskclan/platform | Typed client: dispatch intents, track events, check entitlements, discover registries |
@taskclan/hivemind | Agentic layer: run named agents, shared memory, recall insights, recommend defaults |
@taskclan/hivemind-react | React UI: <HivemindProvider>, <HivemindChat>, useChat, <AgentCard>, <UsageBadge> |
30-minute path
import { createPlatformClient } from '@taskclan/platform';
const client = createPlatformClient({
baseUrl: 'https://api.hivemindos.taskclan.com',
product: 'my-app',
getToken: async () => process.env.HIVEMIND_TOKEN!,
});
// every backend capability is one call:
const result = await client.dispatchIntent('your.intent.name', { /* input */ });→ Quickstart · → Concepts · → Reference
How extension works
Hivemind OS is designed to be extended, not just consumed. Three ways to build with it:
1. Dispatch intents from any service
The platform is HTTP + typed JSON. Any service — your Node backend, a Python
worker, a Rails monolith, a Cloudflare Worker, a bash script with curl —
can call it. The SDKs are for convenience; the wire protocol is
universal.
2. Build your own agents, skills, and tools
Author an agent (a persona-scoped grouping of intents), a skill (a reusable
capability), or a tool (an external connector — Gmail, Slack, GitHub, your own
API). Register it once; the runtime handles scheduling, observability, retries,
and surfacing it in listAgents() / listSkills().
3. Publish to the marketplace
Ship your agents, workflows, prompts, templates, plugins, or tools to the marketplace. Free or paid installs (80/20 creator/platform split on paid). First-party listings are derived live from the registries, so day-one population is real.
Provider-agnostic by design
Hivemind doesn’t lock you into a single AI provider, a single payment processor, or a single deploy target.
- AI providers: the AI gateway routes across OpenAI, Anthropic, OpenRouter, DeepSeek, and Google — with automatic provider failover when one is down.
- Payments: the unified billing ledger spans Stripe, Paystack, RevenueCat, and “manual” (for custom flows).
- Infrastructure: runs on Heroku today; the deploy target is abstracted (multi-cloud roadmap).
Want to swap one out? You write one provider adapter, not a refactor.
Authentication
Today — you authenticate with a Supabase user JWT (
HIVEMIND_TOKEN), the same session you get from signing in at /cloud/signin. For local dev, any Supabase access token works.Roadmap — a first-class developer API-key system (per-key issuance, scoping, rate-limiting, usage attribution) is in flight. Today’s JWT path is fine for dev and early production; production deployments at scale should track the API-key release.
Pricing
Free during the beta. When pricing lands, your existing apps keep working on a free tier; commercial limits will be announced with notice. The Cloud billing surface (/cloud/billing) already shows the eventual plan shape (Free / Pro / Scale / Enterprise).
Verify it yourself
Every claim on this page is checkable on production right now.
- Live products: gamenova.io (flagship, 2,374 users)
- The platform: every API hits one endpoint —
POST /api/hive/v1/intent - Marketplace API (public, no auth):
GET /api/marketplace/v1/listings - Hivemind Cloud: api.hivemindos.taskclan.com/cloud
- Run on the hive: email daniel@taskclan.com for an admin invite to see the live registry, scorecard, and cost ledger
Hivemind OS is a Taskclan product. taskclan.com · hivemindos.taskclan.com