CUSTOMER PLATFORM
Sign in to VARION
Access your complete customer dashboard with your email and password. Manage provider keys, gateway keys, usage, plans, teams, support and measured savings.
Access your complete customer dashboard with your email and password. Manage provider keys, gateway keys, usage, plans, teams, support and measured savings.
Check your inbox or resend the verification email.
Usage and billing status.
Test first, then connect production.
Operational and usage information only. Prompt content is not shown.
| Time | Provider/model | Status | Total input | Provider input | Context avoided | Reduced-cost input | Actual cost | Net saved | Mode / route | Integrity | Fallback |
|---|
Run compatibility and savings previews without exposing VARION’s proprietary engine methods.
Paste a prompt to run a general compatibility preview.
Results will appear here.
Paste an Anthropic/OpenAI-style messages JSON array for a compatibility preview.
Results will appear here.
Paste a tools array and verify every schema is preserved in Zero-Loss Cache mode.
Results will appear here.
Choose a TXT, JSON or Markdown file. VARION reports its token estimate without rewriting it.
Results will appear here.
Estimate tokens without making any AI call.
Results will appear here.
Confirm that whitespace and wording remain byte-for-byte unchanged.
Results will appear here.
Estimate potential provider savings from your expected eligible traffic rate.
Results will appear here.
Confirm the dashboard connection and see exactly what stays private.
Results will appear here.
Uses the encrypted Anthropic key saved in API keys and returns a broad financial result without exposing internal engine details.
Measured cost and savings results will appear here.
Only metrics are stored—never provider keys, prompts or generated answers.
| Time | Provider/model | Verdict | Baseline input | Actual input | Reduced-cost input | Reduction | Verified |
|---|
Connect Claude Code to VARION and review measured provider costs and savings in your dashboard.
Create a separate VARION API key in API keys, then copy it once.
No ZIP, proxy, npm package or installer is required.
set "ANTHROPIC_BASE_URL=https://api.varion.tech" set "ANTHROPIC_AUTH_TOKEN=YOUR_VARION_API_KEY" set "ANTHROPIC_API_KEY=" claude
Close your terminal completely, open a new terminal and start Claude Code normally.
claude
Connected is shown only after a real successful Claude Code request reaches Anthropic through VARION.
No successful Claude Code request yet.
Use a VARION key as the SDK key. Save the OpenAI provider key encrypted in API keys.
Use the OpenAI-compatible VARION base URL with your VARION gateway key.
OpenAI and Anthropic support Zero-Loss Cache. Combined Context + Cache is available to every customer account by default. Savings vary by workload.
Zero-Loss Cache is the strict zero-risk mode: same prompt content, provider and model, with fresh generation and provider-native caching. Auto and Maximum Savings remain optional higher-savings modes with original-request fallback.
The selected mode is the account default. A request may override it with the X-VARION-Mode header when permitted.
| Mode | Context | Provider cache | Fallback |
|---|---|---|---|
| Auto | Reduced only after safety approval | Yes | Original cache-only request |
| Maximum Savings | Tenant-configured beta limits | Yes | Original cache-only request |
| Zero-Loss Cache | Unchanged | Yes | Original direct request |
| Disabled | Unchanged | No | Not applicable |
Use one of these values without changing your request body.
X-VARION-Mode: auto X-VARION-Mode: maximum_savings X-VARION-Mode: zero_loss_cache X-VARION-Mode: disabled
The Proof Lab runs direct cold, direct warm, VARION cold and VARION warm calls. It compares prompt-content hashes, provider/model identity, distinct provider response IDs, cache usage and measured costs.
Use the no-install browser verifier and download a complete evidence ZIP.
Run the comparison from your own computer. The verification source is public; the proprietary VARION engine remains private.
| Protection | Zero-Loss Cache |
|---|---|
| Prompt rewriting | Never |
| Context deletion | Never |
| Semantic response reuse | Never |
| Model or provider switching | Never |
| Stored-response substitution | Never |
| Provider-native cache metadata | May be added |
| New provider generation | Yes |
Use separate VARION gateway keys for devices, environments and team applications.
The full key will appear once.
Only key prefixes are displayed.
| Name | Prefix | Environment | Last used |
|---|
Keys are encrypted at rest and never displayed again. They are excluded from request logs.
Each plan includes a monthly processed-input-token allowance. One successful customer input token deducts one token; failed provider requests deduct nothing.
| Time | Type | Tokens | Balance | Reference |
|---|
Invitation link appears here if email is not configured.
| Name | Role | Status |
|---|
Open a ticket to read replies and continue the conversation.
This ticket is closed. Replying will reopen it.