50 bps. Every frontier model. One bill.

Every frontier model. 90% less markup.

The same models, the same routing, a tenth of the platform fee. Every frontier provider on one balance — chat, API, and 100+ tools.

50 bps
Platform fee on tokens
400+
Models, every frontier provider
50–80%
Spend cut by the auto-router
$0
Per-seat fees. Ever.
Skinny margins, by design

50 bps on tokens. A tenth of the standard.

Most enterprise AI platforms run a 5–10% markup on every token. We charge 50 bps — enough to keep the lights on, not enough to be a line item in your budget. We make our real margin on the agent infrastructure layer above the tokens.

If your team spends $1M/year on tokens:
Platform Platform fee Annual cost over provider
ATXP Chat 0.50% $5,000
Standard gateway 5.5% $55,000
ChatGPT Business (1,000 seats) $30/seat/mo $360,000

$50,000 saved on a $1M workload — a 90% lower platform fee than the market rate. Scales linearly: at $10M of tokens, that's $500K back in your budget.

One platform

Every model. Every tool. One balance, one bill.

ATXP is the agent-native platform underneath your AI strategy — the chat UI your team actually uses, the API your engineers build on, and the tools your agents call to do real work. Not just a router.

Auto-router, by task

We route each message to the model that's best for it — quality and cost. Teams typically see 50–80% lower spend without changing what they ask.

🧰

Tools, not just tokens

Image, video, voice, web search, SMS, email. Your team and your agents call them from the same balance, billed transparently.

🪞

Chat UI included

Most gateways are API-only plumbing. ATXP ships the surface employees actually use — plus the API your engineers build on. Same platform, same pricing.

📊

Spend controls

Daily, weekly, monthly limits. Per-team, per-user, per-model budgets. Activity logs and exportable traces.

🔐

SSO & SAML, included

Google Workspace and Okta out of the box. SAML for the rest. No "Enterprise tier upgrade" required for the basics.

🔁

Automatic failover

Provider rate-limited? Region down? Requests reroute automatically. Your product keeps shipping.

Claude Opus 4.8 GPT-5.5 Pro Gemini 3.1 Pro Grok 4 Llama 4 Mistral Large DeepSeek V3 Qwen 3 + 400 more
Privacy & governance

Sovereign AI, by default.

Your conversations and your code aren't training data. Period. Routing, residency, and retention are policies you set — not surprises we change.

SOC 2 readiness, zero data retention on supported providers, BYOK for the providers you've already procured. Export or delete anything, anytime.

1

Zero training, zero retention

We don't train on your data. ZDR-by-default routing for providers that support it; opt-in for the rest.

2

BYOK supported

Bring your existing Anthropic, OpenAI, or Google contract. Route some workloads through your keys, others through ours.

3

Auditable everything

Per-user, per-team, per-model activity logs. Stream traces to Datadog, S3, or your SIEM of choice.

4

No lock-in

Open data export, open spec. Walk away with everything you put in. Nothing about your AI strategy should depend on us being here in 2030.

Side by side

How we compare.

An honest line-by-line against the alternatives. We win on cost and scope; we lose on brand recognition (for now).

ATXP Chat OpenRouter ChatGPT Business ChatGPT Enterprise
Platform fee on tokens 0.50% 5.5% Bundled in seats Bundled in seats
Per-seat fees $0 $0 $30/seat/mo Contact sales
Annual contract required No No Annual billing Yes
Frontier models 400+ 400+ OpenAI only OpenAI only
Chat UI for employees Yes No (API only) Yes Yes
Auto-router for cost Yes (50–80% savings) Failover only No No
Tools beyond LLMs Image, video, voice, search, SMS No Limited Limited
SSO / SAML Included Enterprise tier only Add-on Yes
Field engineers Included at $50k/mo No No Add-on
Data used for training Never Never Never Never
FDE Program

We send engineers, not account managers.

Forward-deployed engineers who build the AI workflows your team actually needs — then hand them off so you run independently. Auto-included once your account crosses $50k/mo in spend. No procurement call required.

Start free →
Map the 5 workflows where AI moves the most revenue or cost
Build custom agents and integrations on your stack
Deploy managed agents that run autonomously on your behalf
Train your team to keep shipping after we leave
Priority support and direct Slack channel
FAQ

The honest questions.

How does the 50 bps fee work?

We pass through Anthropic, OpenAI, Google, xAI, and others at their published rates, plus a flat 50 basis points (0.5%) platform fee on tokens. Volume discounts we negotiate get passed through too. No hidden middle tier.

How is this different from OpenRouter?

Different layer of the stack. OpenRouter is API-only routing infrastructure. ATXP ships the full surface — chat UI for employees, API for engineers, and tools beyond LLMs (image, video, voice, search) on one balance. Our platform fee comes in at 50 bps vs the 5.5% standard.

How is this different from ChatGPT Business / Enterprise?

No per-seat fees, no annual contract, every frontier model (not just OpenAI), and auto-routing that cuts spend 50–80%. Same data privacy guarantees.

What about SOC 2 and residency?

SOC 2 readiness in progress, in-region routing for supported providers. Email security@atxp.ai for the current security pack.

Can we BYOK?

Yes. Route some workloads through your existing Anthropic / OpenAI / Google contracts, others through ours. Useful when you've pre-paid commitments to burn down.

What's the smallest contract size?

None. Pay-as-you-go from $0. The FDE program kicks in at $50k/mo spend. Enterprise procurement (invoicing, POs, custom DPAs) available at any volume.

Can we try before signing anything?

Yes — Consumer and Business are free to start, same product. Run a real workload through it. Upgrade to Enterprise when you want FDE coverage or procurement terms.

Ready to put your whole company on one balance?

No procurement call, no annual contract, no minimum. Spin up a workspace and route real volume through it today.