One inference network, priced in rand.
The VZN Kernel routes every message through one decision point: a template on the hot path, a small model when it is unsure, and a larger model only when a job opts in and budget remains.
Cheap answers on the hot path, paid models only when they are worth it.
- The ACK path stays free of LLM calls, so WhatsApp and webhook ingest answer in milliseconds.
- Paid inference is an async sidecar behind allowlists, a per-job budget and a hard daily ceiling.
- Four paid models have verified list prices; lab-free models cost nothing.
Every product, one price list.
Free and open tiers where it makes sense, per-seat plans for the operator tools, and usage billing for inference and agents that matches the Hub meter. All figures are proposed and approved before promotion to main.
Verified list prices, held in rand at R18.5/USD.
DeepSeek V4.1 Flash
Tier 1R5.55 in / R22.20 out per million tokens.
1M token context; off-peak billing is half the peak rate the registry holds
DeepSeek V4 Pro
Tier 2R24.42 in / R73.26 out per million tokens.
1M token context; measured on the coding eval at R7.06 for 58/60 runs
Premium Tier 1
Tier 1R18.50 in / R92.50 out per million tokens.
First-party pricing with cache-read and cache-write rates in the registry
Premium Tier 2
Tier 2R37 in / R185 out per million tokens.
The default when a premium key is set and no model is chosen
List prices were checked 2026-10-16 against the providers. The registry holds the peak rate, so an estimate is never low; set INFERENCE_ZAR_PER_USD to move the exchange rate without a deploy.
Pay for the tokens your agents use, not for seats that sit idle.
Self-hosted
R0 / month
Run the agent loop on your own model, your hardware, R0 per call.
- Agent loop and workflow engine
- APPROVE gate stays on
- Self-hosted model in the registry
Pay as you go
Most commonMetered
Metered like the Hub: tokens in, tokens out, cache reads and batch discounts.
- R5.55 in / R22.20 out per 1M tokens on DeepSeek V4.1 Flash
- Per-job and per-day rand budgets
- The meter's own ledger, nothing estimated
Enterprise
Talk to us
Hosted ops, private models and a support agreement.
- Private deployment
- Billing and compliance review
- Talk to us
The usage units are the Hub meter's own: tokens_in, tokens_out, cache_read_tokens and zar_est.
The design language is open; support and design reviews are not.
Open core
R0 / one-time
React, React Native and the token core, from one MIT-licensed source.
- @vzn-labs/ui, ui-native and ui-core
- Token source in the repo
- The live showcase
Support
Talk to us
Design-system reviews, Figma work and white-label theming.
- A named designer
- Brand reviews
- Talk to us
The packages are built in this repo but not published to npm yet.
One price per person who writes campaigns; every line is fact-checked before it leaves.
Free
R0 / month
Try the ten fact checks on your own campaigns.
- Text campaigns
- Ten checks per line
- Export approved lines
Pro
Most commonR290 / month
Per seat, with the checks and approvals teams actually need.
- Everything in Free
- Team approvals
- Campaign history
Enterprise
Talk to us
More seats, your fonts, your review workflow.
- Unlimited seats
- Custom review rules
- Talk to us
Write once, post to the accounts you connect yourself.
Free
R0 / month
One network, scheduled posts, nothing posts without you.
- One connected network
- Scheduling
- Approval stays on
Pro
Most commonR190 / month
All seven networks from one queue.
- Seven networks
- Queue and scheduling
- Post history
Enterprise
Talk to us
Multi-brand and agency workflows.
- Multiple brands
- Approval roles
- Talk to us
Your desk is an MCP server; scoped keys are the only way in.
Self-hosted
R0 / one-time
Run the desk MCP server yourself with scoped connection keys.
- Streamable HTTP MCP server
- Scoped keys
- Tool allowlist
Team
Most commonR150 / month
Shared connectors and an allowlist per agent.
- Shared connector catalogue
- Per-agent allowlists
- Usage history
Enterprise
Talk to us
OAuth for remote web clients and private connectors.
- OAuth
- Private connectors
- Talk to us
Teach an agent a recipe once, reuse it everywhere.
Free
R0 / month
Personal recipes with validation and an audit trail.
- Create and edit recipes
- Validation
- Audit trail
Team
Most commonR120 / month
Shared recipes your whole team can reuse.
- Shared recipes
- Bind skills to agents
- Audit trail
Enterprise
Talk to us
A private skill store and imported Markdown skills.
- Private store
- Markdown import
- Talk to us
The harness is free; hosted evals are metered per case.
Harness
R0 / one-time
Eight suites and 72 cases, run against reference answers.
- Eight suites
- 72 cases
- Command-line runner
Hosted evals
Most commonMetered
Run the same cases on our hardware, priced per case.
- Per-case metering
- Weekly schedule
- Results you can rerun
Enterprise
Talk to us
A published leaderboard and operator observability.
- Leaderboard
- Observability
- Talk to us
Customer records where agents propose the write and you APPROVE it.
Starter
R350 / month
Per seat, for a small team triaging one inbox.
- Contacts and activities
- APPROVE-gated writes
- One inbox
Growth
Most commonR650 / month
Deals, pipeline and a live connector.
- Deals and pipeline
- Import from external CRMs
- Team approvals
Enterprise
Talk to us
Multiple brands and compliance review.
- Multi-brand
- Compliance posture
- Talk to us
Stock, orders and invoices, with agents doing the work.
Starter
R450 / month
Per seat, for a small operation.
- Stock and orders
- APPROVE-gated writes
- MCP connectors
Growth
Most commonR850 / month
Invoices and multi-warehouse stock.
- Invoices
- Multi-warehouse
- Team approvals
Enterprise
Talk to us
Custom connectors and compliance review.
- Custom connectors
- Compliance posture
- Talk to us
The typed client is free; the API it calls is metered like the Hub.
Free
R0 / one-time
A typed TypeScript client over the routes developers already call.
- Typed client
- Desk token or scoped key
- Same 401/403 auth lock
Usage
Most commonMetered
Pay for Kernel API usage at the registry's verified rates.
- Inference at list price
- Rand budgets
- The meter's own ledger
Enterprise
Talk to us
Private endpoints and support.
- Private endpoints
- Support
- Talk to us
Bring your own keys and run it free; pay for team features later.
Free
R0 / one-time
The coding agent with BYOK, plan mode and undo.
- Any OpenAI-compatible model
- Free providers with failover
- Hard rand budget
Pro
Most commonR99 / month
Shared workspaces and per-hunk approve for a team.
- Shared workspaces
- Per-hunk approve
- Command sandbox
Enterprise
Talk to us
Company policy and provider contracts.
- Policy controls
- Provider contracts
- Talk to us
Costs are held in USD and shown in rand by default, with a hard budget that stops the run before it overspends.
Browse and install industry packages; paid packages are per seat.
Free
R0 / one-time
Install open industry packages.
- Open packages
- Skills and control cards
- Local install
Pro
Most commonR150 / month
Paid packages and updates, per seat.
- Paid packages
- Updates and rollback
- Install history
Enterprise
Talk to us
A private marketplace for your company.
- Private marketplace
- Governance
- Talk to us
The phone and desktop clients are included; the local engine is optional.
Free
R0 / one-time
The phone web build and the desktop shell.
- Jobs board on a phone
- APPROVE from the phone
- Desktop shell
Local engine
Most commonR149 / month
Run the Rust engine locally with your own keys.
- Local engine
- OS-keychain keys
- Offline sessions
Enterprise
Talk to us
Signed installers and managed updates.
- Signed installers
- Auto-update
- Talk to us