AI marketplace · governed by default
One place
for all
your AI.
Discover agents, pay one bill, see what every dollar bought.
Already have a workspace? Sign in
Models
AI agents
One bill · August 2026
Illustrative figures — not a customer account
7 providers · all your AI agents · one gateway · per-request cost
- AI providers, one gateway
- 0
- Apps you can install today
- 0
- Of requests costed and logged
- 0%
- Provider SDKs in your codebase
- 0
AI providers, one gateway
Apps you can install today
Of requests costed and logged
Provider SDKs in your codebase
The problem
Your company already runs on AI. You just can’t see it.
Most businesses under 200 people are paying four or five AI vendors separately, with no shared view of who is spending what, and no way to stop paying for the tool that isn’t working.
Today
Five vendor bills, five logins, five renewal dates.
One bill. Every provider and every agent on it, itemised.
No idea which team, person, or project is spending what.
Spend attributed per employee, per department, per request.
An agent underperforms, but you already paid that vendor.
Swap it out. The budget stays with you, not the vendor.
“Don’t paste client data into AI” is a Slack message nobody enforces.
Policy runs on every request, before the model sees it.
The marketplace
Agents and governance your team can install today.
Every capability is an installable app. Turn it on for your workspace — no redeploy, no release to wait for, no engineer required.
Planned — not yet available
- Redundant AI Reruns Stopper
Apps are built by us and by third-party developers. Build and list your own.
How control is enforced
One gateway in front of everything.
Every AI request your company makes takes the same four steps. Humans write the policy; the platform carries it out, request by request.
01
Request
An employee, an agent, or a coding tool asks for something.
02
Policy
Your rules run first — who, which model, how much, what data.
Set by your admins
03
Provider
Routed to the right model across seven providers.
04
Ledger
Real provider-reported cost recorded against the request.
Spend ceilings
Dollar caps per employee, group, position or department, checked before the request runs rather than discovered on the invoice.
Model routing
Decide which models each role can reach. A denial fails loudly and gets logged — it is never quietly downgraded to something cheaper.
Data rules
Block file uploads, image generation, or prompts carrying client data, per company or per department.
Full audit trail
Every request, permission change, login and denial is recorded and queryable.
Routes to OpenAI, Anthropic, Gemini, Azure OpenAI, DeepSeek, Grok and Ollama. Your business code never touches a provider SDK, so adding a provider changes nothing you have already built.
What you see
Every request, costed — including your agents.
Not an estimate from a token counter. The real, provider-reported cost of each call, attributed to the person, the project, and the agent that made it.
| Source | Model routed | Requests | Share | Spend |
|---|---|---|---|---|
| Models | ||||
| Anthropic | claude-opus-5 | 18,402 | 39.5% | 412.80 |
| OpenAI | gpt-5.6-terra | 11,097 | 18.0% | 188.20 |
| Gemini | gemini-3.6-flash | 4,881 | 7.1% | 74.15 |
| Ollama | self-hosted | 2,140 | 0.0% | 0.00 |
| AI agents | ||||
| Claude Code | claude-opus-5 | 9,634 | 20.6% | 214.60 |
| Marketing Analyst | claude-sonnet-5 | 3,180 | 9.2% | 96.40 |
| Business Assistant | gpt-5.6-luna | 2,904 | 5.6% | 58.00 |
| Total | 1,044.15 | |||
Illustrative figures — not a customer account
The same ledger reconciles each provider’s own billing total against what we measured, and leads with the difference — the spend that never went through the gateway at all.
Getting started
An afternoon, not a project.
There is nothing to deploy and nothing to migrate. Your team keeps using the tools they already use.
01
Create your workspace
One admin, one company. No sales call, no procurement review, no implementation partner.
02
Connect your providers
Bring your existing OpenAI, Anthropic or Gemini keys, or draw on the platform balance and skip the vendor accounts entirely.
03
Install what you need
Spend caps, data rules, agents. Enable per company or per department; changes take effect on the next request.
04
Point your team at it
The web app, the API, or the VS Code extension that routes an unmodified Claude Code through the same governed gateway.
For AI agent builders
Distribution, not another app directory.
If you are building a vertical AI agent, your hardest problem is reaching businesses that have a real use case and real budget. That is the side of this marketplace we are building for you.
Buyers with budget already in hand
Your listing is in front of businesses that have already connected AI spend, not cold traffic evaluating whether they need AI at all.
Your code runs on your infrastructure
Nothing ships to us. We call your endpoint, or you call a narrow capability API scoped to exactly what the customer granted.
Billing and metering handled
Charge the customer or absorb the cost yourself. Usage is metered per request either way, and you get aggregate performance data for your app.
A sandbox before anyone sees you
Every unapproved app can only ever reach one reserved synthetic tenant. You build against real APIs without touching a real customer.
Trust
What’s true today, including the parts that aren’t flattering.
You are being asked to route your company’s AI through us. Here is exactly what we do and do not hold — in the same words as the evidence pack we send to procurement.
Not held
SOC 2 Type II
Not held. Controls described in our evidence pack are implemented and evidenced there, but have not been audited by a third party.
ISO 27001
Not held.
Third-party penetration test
No third-party penetration test has been commissioned to date.
In place
Encryption at rest
Provider credentials and captured conversation content are encrypted at rest (Fernet/AES-256) with keys held outside the database.
Encryption in transit
TLS on every external interface.
Tenant isolation
Enforced at the ORM layer: every tenant-scoped model reads through a manager bound to the request's company, with a single greppable cross-tenant escape hatch used only by platform-operations sweeps.
Authorization as data
Capabilities are permission codenames attached to roles, not role checks written into business logic.
Full detail, subprocessors and a downloadable evidence pack on the security page.
Pricing
No subscription. We earn when you spend.
There is no monthly fee, no seat count and no minimum. You pay for the AI you actually use, at cost plus a flat rate — which is the only reason you can trust us to show you the cost.
You pay
cost + TBD%
on AI usage, nothing else
Every provider and every agent on one bill. No seats, no tiers, no minimum.
Vendors pay
TBD%
revenue share, no listing fee
Reach businesses with budget already connected and actively browsing.
Free, always
0.00
the whole governance layer
Spend visibility, audit trail, limits and the marketplace itself cost nothing.
What does it cost to start?
Nothing. There is no monthly fee, no seat count and no minimum. You pay a flat rate on top of the AI usage you actually consume, and the governance layer — spend visibility, audit trail, limits and the marketplace itself — costs nothing.
Can we use our own provider accounts?
Yes. Bring your existing OpenAI, Anthropic, Gemini, Azure OpenAI, DeepSeek, Grok or Ollama credentials and we govern and meter the traffic without reselling you anything. The take rate applies to usage drawn from the platform balance.
How is our data isolated from other companies?
Structurally, not by convention. Every tenant-scoped model auto-scopes reads and writes to your company through a request-bound context variable rather than a filter someone has to remember to add.
What happens to our data if we leave?
Conversation content is retained for 60 days after deletion and then hard-deleted by a scheduled sweep. You can export your workspace at any time before that.