AGICY Copperway · EU-sovereign AI gateway
Frontier AI without surrendering your data
Frontier models through sovereign copper.
Frontier models through sovereign copper. PII masked in an AES-encrypted vault on EU compute — in-memory rehydration only, no raw prompts to US upstream. One OpenAI-compatible URL for the models we list — not a 500-model marketplace.
- PII vault
- EU residency
- AES encrypted at rest
- In-memory rehydration
- PII masked pre-route
Claude
OpenAI
Gemini
Grok
Mistral
DeepSeek
Kimi
Qwen
Zhipu
MiniMax
- PII vault
- EU residency
- AES vault
- OpenAI-compatible
Gateway endpoint https://proxy.agicy.com/v1Drop-in OpenAI SDK · audited demand model: Bank of Cyprus 148.3B tokens/yr
Routing blueprint · Cyprus EU compute path
Open elevation sheet →If you already use OpenRouter
US credit marketplace versus an EU path with a PII vault. OpenRouter publishes 5.5% on credits and 5% BYOK after allowance — no markup on model list. Copperway SRA intro is €0.10/M in + €0.10/M out.
If you already use OpenRouter, you have a wide catalog and auto-routing. Copperway is the other job: an OpenAI-compatible EU gateway with a PII vault. Five percent of a $15/M frontier output rate is $0.75/M; SRA output is €0.10/M (~$0.109/M at planning FX 0.92). Semantic compression can shrink billed input; OpenRouter prompt caching is a different lever. Catalog width is theirs. Flat SRA plus PII-before-upstream are ours.
OpenRouter’s live homepage states “200T+ Monthly Tokens” — monthly, not all-time. OpenRouter is hosted SaaS (Free, pay-as-you-go, Enterprise), not open source.
Homepage volume, checked 2026-08-19: 200T+ Monthly Tokens (monthly). Source: https://openrouter.ai/.
| Model | List in | List out | 5% BYOK take in | 5% BYOK take out | 5.5% credit fee in | 5.5% credit fee out | Copperway SRA in | Copperway SRA out | SCE input saving (46%) | SRA vs 5% BYOK out |
|---|---|---|---|---|---|---|---|---|---|---|
| $150.00/M | $600.00/M | $7.50/M | $30.00/M | $8.25/M | $33.00/M | €0.10/M ($0.109/M) | €0.10/M ($0.109/M) | $69.00/M | $29.891/M cheaper | |
| $30.00/M | $180.00/M | $1.50/M | $9.00/M | $1.65/M | $9.90/M | €0.10/M ($0.109/M) | €0.10/M ($0.109/M) | $13.80/M | $8.891/M cheaper | |
| $30.00/M | $180.00/M | $1.50/M | $9.00/M | $1.65/M | $9.90/M | €0.10/M ($0.109/M) | €0.10/M ($0.109/M) | $13.80/M | $8.891/M cheaper | |
| $30.00/M | $150.00/M | $1.50/M | $7.50/M | $1.65/M | $8.25/M | €0.10/M ($0.109/M) | €0.10/M ($0.109/M) | $13.80/M | $7.391/M cheaper | |
| $21.00/M | $168.00/M | $1.05/M | $8.40/M | $1.155/M | $9.24/M | €0.10/M ($0.109/M) | €0.10/M ($0.109/M) | $9.66/M | $8.291/M cheaper | |
| $15.00/M | $120.00/M | $0.75/M | $6.00/M | $0.825/M | $6.60/M | €0.10/M ($0.109/M) | €0.10/M ($0.109/M) | $6.90/M | $5.891/M cheaper | |
| $20.00/M | $80.00/M | $1.00/M | $4.00/M | $1.10/M | $4.40/M | €0.10/M ($0.109/M) | €0.10/M ($0.109/M) | $9.20/M | $3.891/M cheaper | |
| $15.00/M | $75.00/M | $0.75/M | $3.75/M | $0.825/M | $4.125/M | €0.10/M ($0.109/M) | €0.10/M ($0.109/M) | $6.90/M | $3.641/M cheaper | |
| $15.00/M | $60.00/M | $0.75/M | $3.00/M | $0.825/M | $3.30/M | €0.10/M ($0.109/M) | €0.10/M ($0.109/M) | $6.90/M | $2.891/M cheaper | |
| $30.00/M | $60.00/M | $1.50/M | $3.00/M | $1.65/M | $3.30/M | €0.10/M ($0.109/M) | €0.10/M ($0.109/M) | $13.80/M | $2.891/M cheaper |
| $10.00/M | $50.00/M | $0.50/M | $2.50/M | $0.55/M | $2.75/M | €0.10/M ($0.109/M) | €0.10/M ($0.109/M) | $4.60/M | $2.391/M cheaper | |
| $10.00/M | $50.00/M | $0.50/M | $2.50/M | $0.55/M | $2.75/M | €0.10/M ($0.109/M) | €0.10/M ($0.109/M) | $4.60/M | $2.391/M cheaper | |
| $10.00/M | $50.00/M | $0.50/M | $2.50/M | $0.55/M | $2.75/M | €0.10/M ($0.109/M) | €0.10/M ($0.109/M) | $4.60/M | $2.391/M cheaper | |
| $5.00/M | $30.00/M | $0.25/M | $1.50/M | $0.275/M | $1.65/M | €0.10/M ($0.109/M) | €0.10/M ($0.109/M) | $2.30/M | $1.391/M cheaper | |
| $3.00/M | $15.00/M | $0.15/M | $0.75/M | $0.165/M | $0.825/M | €0.10/M ($0.109/M) | €0.10/M ($0.109/M) | $1.38/M | $0.641/M cheaper |
Modeled comparison. Not a quote. SCE 46% is modeled and fail-open. SRA intro is subsidized. Campus is pre-construction; executed offtake is zero. Terms · Gateway pricing disclaimer.
Copperway vs LLM gateways — what we ship · CLOUD Act / EU vs US marketplace · LiteLLM vs Portkey vs Copperway · Try Copperway Playground
Next: How traffic routes →
How traffic routes
Five deterministic stages between your application and upstream model providers — the Copperway sovereign gateway pipeline for privacy, semantic compression, and auditable routing. Orchestration expands Stage 04 to fan out across frontier APIs and sovereign open-weight models.
Model agnostic · orchestration
One OpenAI-compatible Copperway endpoint — frontier APIs (Claude Sonnet 4.6, GPT-4.1, Gemini 2.5, Grok 4.5) and sovereign open-weight models on the same pipeline. Orchestration engine — Rust microservice (roadmap) — sits after PII mask and SCE compression, before upstream fan-out: auto-switch by policy, parallel replies per topic, or agent planners that dispatch code to small models and reasoning to frontier.
Sovereign open model first — escalate to frontier only when confidence is low. ~85% cost cut vs frontier-only (RouteLLM benchmark, lmsys.org 2024).
Same compressed prompt to N models. Race for latency, score for quality, or return all for ensemble evaluation.
Planner dispatches sub-tasks: code → small sovereign model, reasoning → frontier. Multi-step without client-side routing.
Policy by cost cap, latency SLA, PII tier, and EU residency — workspace rules pick the mode per request.
Open-weight Llama and Mistral on Tenstorrent fleet at near-zero marginal vs frontier API. Frontier routing bills at Standard Copperway (12.4–17.6% blended rate saving) or SRA intro subsidized flat rate — see sheet 05. Orchestration hot path: Rust microservice on Cyprus sovereign fleet; Python offline planner for agents and routing policy cache — SDK schemas open, engine proprietary. See docs/GATEWAY_ORCHESTRATION_PROPOSAL.md.
Why route through Copperway
Direct API calls expose regulated data and bill on raw prompt length. Copperway adds EU sovereignty, PII vaulting, and auditable compression — 46% fewer input tokens billed → 12.4–17.6% blended rate saving → up to ~54% total bill reduction — without forking your stack. Gateway pricing disclaimer →
Two Copperway tiers — sovereign routing included
SRA intro: €0.10/M input + €0.10/M output (subsidized). Standard Gateway: 12.4–17.6% blended rate saving via post-SCE billing + €0.15/M platform fee. PII masking and EU residency on both.
Direct OpenAI API
- Full prompt text sent to US-domiciled provider
- No deterministic PII stripping before upstream
- Full token billing on raw prompt length
- No in-flight semantic compression layer
- Compliance review blocks production rollout
AGICY Copperway
- OpenAI-compatible — swap gateway base URL only
- PII masked and vaulted before upstream routing
- SCE bills fewer post-compression tokens upstream
- Responses rehydrated in-memory on delivery
- EU-incorporated path with a PII vault — campus is pre-construction
Frontier API supplier rankings
AGICY's estimated Copperway rates by tier — SRA intro (subsidized flat) or Standard Gateway (post-SCE compressed pass-through + platform fee) — compared against official list prices for 30 frontier models. 46% fewer input tokens billed → 12.4–17.6% blended rate saving → up to ~54% total bill reduction. Not official reseller pricing →
Choose your pricing tier
Two clear rates — no mixed headlines. SRA intro is subsidized; Standard Gateway is the sustainable pass-through rate for production.
Standard Gateway — 12.4–17.6% vs direct API
Standard Gateway: compressed upstream pass-through (46% fewer input tokens billed) plus €0.15/M platform fee. SRA intro flat €0.10/M in + €0.10/M out is subsidized — select SRA tier above to compare.
| Model | Direct blended | Standard Gateway est. | Savings | Reserve |
|---|---|---|---|---|
| Claude Sonnet 4.6 | $6.00/M€5.52/M | €4.72/Mpass-through + platform fee | $0.87/M · 14.5% | |
| Claude Opus 4.8 | $10.00/M€9.20/M | €7.76/Mpass-through + platform fee | $1.56/M · 15.6% | |
| GPT-5.5 | $11.25/M€10.35/M | €8.91/Mpass-through + platform fee | $1.56/M · 13.9% | |
| Gemini 3.1 Pro | $5.63/M€5.17/M | €4.53/Mpass-through + platform fee | $0.70/M · 12.4% | |
| Grok 4.5 | $3.00/M€2.76/M | €2.27/Mpass-through + platform fee | $0.53/M · 17.6% |
AGICY pricing — our rates
Two tiers: SRA intro — €0.10/M input + €0.10/M output (subsidized via sovereign allocation). Pass-through — compressed upstream list + €0.15/M platform fee (12.4–17.6% vs direct on flagship models).
| Service | AGICY price | vs direct | Savings |
|---|---|---|---|
| Gateway platform fee (sovereign routing) | €0.15/M | — (not charged by upstream) | Covers proxy ops + PPA margin |
| Frontier API input — Claude Sonnet 4.6 (compressed pass-through) | $1.62/M upstream + €0.15/M platform | $3.00/M | ~46% fewer input tokens billed |
| Frontier API output — Claude Sonnet 4.6 (upstream pass-through) | $15.00/M + €0.15/M platform | $15.00/M | 0% — billed at upstream list |
| Frontier API blended (3:1 input:output mix) | ~€4.72/M all-in est. | ~€5.52/M direct list | ~14.5% vs direct (modeled) |
| SRA intro flat rate (subsidized) | €0.10/M in + €0.10/M out | Modeled COGS ~€4.68/M | Promotional — sovereign cross-subsidy |
| Sovereign compute — energy marginal (on-campus) | €0.43/M COGS basis | ~€2+/M cloud GPU-equivalent | Up to 8× vs hyperscaler (plan doc) |
AGICY Gateway all-in rate: €0.10/M input + €0.10/M output (€0.20/M per 1M in + 1M out). Modeled frontier COGS ~€4.68/M vs retail €0.10/M at 3:1 mix — subsidized via sovereign allocation. Includes PII masking, EU routing, and ~46% SCE input compression.
List prices from provider APIs / LLM Stats reference · asOf 2026-07-10. Direct blended uses 75% input / 25% output weighting. AGICY Gateway SRA intro: €0.10/M input + €0.10/M output — subsidized via sovereign compute allocation. Standard Gateway: compressed upstream pass-through + €0.15/M platform fee. 46% input compression reduces billed input tokens. AGICY is not an official reseller of upstream providers. Full gateway pricing disclaimer → · Sovereign compute tiers →
| # | Model | Provider | Context · speed | List in / out | Direct blended | AGICY Standard est. | AGICY SRA intro | Savings % | Actions |
|---|---|---|---|---|---|---|---|---|---|
| 1 | Claude Fable 5 claude-fable-5 | Anthropic | 1M20 tok/s | $10.00 / $50.00 | $20.00/M€18.40/M | €15.38/Mpass-through + platform | €0.10/Mflat SRA intro | −16.4%SRA intro −99.5% | |
| 2 | Claude Opus 4.8 claude-opus-4-8 | Anthropic | 1M48 tok/s | $5.00 / $25.00 | $10.00/M€9.20/M | €7.76/Mpass-through + platform | €0.10/Mflat SRA intro | −15.6%SRA intro −98.9% | |
| 3 | Claude Opus 4.7 claude-opus-4-7 | Anthropic | 1M34 tok/s | $5.00 / $25.00 | $10.00/M€9.20/M | €7.76/Mpass-through + platform | €0.10/Mflat SRA intro | −15.6%SRA intro −98.9% | |
| 4 | Claude Opus 4.6 claude-opus-4-6 | Anthropic | 1M37 tok/s | $5.00 / $25.00 | $10.00/M€9.20/M | €7.76/Mpass-through + platform | €0.10/Mflat SRA intro | −15.6%SRA intro −98.9% | |
| 5 | Claude Opus 4.5 claude-opus-4-5 | Anthropic | 200K35 tok/s | $5.00 / $25.00 | $10.00/M€9.20/M | €7.76/Mpass-through + platform | €0.10/Mflat SRA intro | −15.6%SRA intro −98.9% | |
| 6 | Claude Sonnet 5 claude-sonnet-5 | Anthropic | 1M75 tok/s | $3.00 / $15.00 | $6.00/M€5.52/M | €4.72/Mpass-through + platform | €0.10/Mflat SRA intro | −14.5%SRA intro −98.2% | |
| 7 | Claude Sonnet 4.6 claude-sonnet-4-6 | Anthropic | 200K40 tok/s | $3.00 / $15.00 | $6.00/M€5.52/M | €4.72/Mpass-through + platform | €0.10/Mflat SRA intro | −14.5%SRA intro −98.2% | |
| 8 | Claude Sonnet 4.5 claude-sonnet-4-5 | Anthropic | 200K42 tok/s | $3.00 / $15.00 | $6.00/M€5.52/M | €4.72/Mpass-through + platform | €0.10/Mflat SRA intro | −14.5%SRA intro −98.2% | |
| 9 | Claude Haiku 4.5 claude-haiku-4-5 | Anthropic | 200K120 tok/s | $1.00 / $5.00 | $2.00/M€1.84/M | €1.67/Mpass-through + platform | €0.10/Mflat SRA intro | −9.1%SRA intro −94.6% | |
| 10 | GPT-5.5 gpt-5.5 | OpenAI | 1.1M37 tok/s | $5.00 / $30.00 | $11.25/M€10.35/M | €8.91/Mpass-through + platform | €0.10/Mflat SRA intro | −13.9%SRA intro −99.0% | |
| 11 | GPT-5.4 gpt-5.4 | OpenAI | 1M231 tok/s | $2.50 / $15.00 | $5.63/M€5.17/M | €4.53/Mpass-through + platform | €0.10/Mflat SRA intro | −12.4%SRA intro −98.1% | |
| 12 | GPT-5.4 mini gpt-5.4-mini | OpenAI | 400K175 tok/s | $0.75 / $4.50 | $1.69/M€1.55/M | €1.47/Mpass-through + platform | €0.10/Mflat SRA intro | −5.7%SRA intro −93.6% | |
| 13 | GPT-5.2 gpt-5.2 | OpenAI | 400K38 tok/s | $1.75 / $14.00 | $4.81/M€4.43/M | €4.02/Mpass-through + platform | €0.10/Mflat SRA intro | −9.2%SRA intro −97.7% | |
| 14 | GPT-5.1 gpt-5.1 | OpenAI | 400K336 tok/s | $1.25 / $10.00 | $3.44/M€3.16/M | €2.92/Mpass-through + platform | €0.10/Mflat SRA intro | −7.8%SRA intro −96.8% | |
| 15 | GPT-5.4 nano gpt-5.4-nano | OpenAI | 400K280 tok/s | $0.20 / $1.25 | $0.46/M€0.43/M | €0.51/Mpass-through + platform | €0.10/Mflat SRA intro | +20.2% vs directSRA intro −76.5% | |
| — |  AGICY CopperwayLive EU gateway · PII mask · SCE compression · campus pre-construction | Upstream list (USD) | 75/25 blended direct | Compressed pass-through + platform feeOption B Standard | €0.10/MOption A SRA flat | 12.4–17.6% blended vs direct API | |||
AGICY Gateway all-in rate: €0.10/M input + €0.10/M output (€0.20/M per 1M in + 1M out). Modeled frontier COGS ~€4.68/M vs retail €0.10/M at 3:1 mix — subsidized via sovereign allocation. Includes PII masking, EU routing, and ~46% SCE input compression.
List prices from provider APIs / LLM Stats reference · asOf 2026-07-10. Direct blended uses 75% input / 25% output weighting. AGICY Gateway SRA intro: €0.10/M input + €0.10/M output — subsidized via sovereign compute allocation. Standard Gateway: compressed upstream pass-through + €0.15/M platform fee. 46% input compression reduces billed input tokens. AGICY is not an official reseller of upstream providers. Standard tier savings can read negative on ultra-low list models (e.g. nano/flash SKUs) where the €0.15/M platform fee exceeds compressed pass-through — SRA intro flat rate is the better fit for those workloads. Full gateway pricing disclaimer →
Reserve gateway capacity via SRA
Production routing at compressed upstream pass-through + €0.15/M platform fee. SRA intro subsidized flat rate available for sovereign capacity customers.
Lock in subsidized SRA intro rates
€0.10/M input + €0.10/M output while sovereign compute allocation lasts. Standard Gateway applies post-intro at compressed pass-through + €0.15/M.
Drop-in integration
Point your OpenAI SDK at the Copperway gateway base URL. Authenticate with your API secret — not your upstream key. Production-ready for regulated workloads in banking, telecom, and public sector.
curl https://proxy.agicy.com/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer YOUR_API_SECRET" \
-d '{
"model": "gpt-4o-mini",
"messages": [{"role": "user", "content": "Summarise this contract clause."}],
"stream": false
}'Use your API secret as the bearer token. Rotate it in the operator console after sign-in.
| Endpoint / mode | Status | Notes |
|---|---|---|
/v1/chat/completions | Supported | Primary path — mask, compress, route, rehydrate |
/v1/images | Supported | Routes to Gemini image generation when configured; prompt is PII-masked |
/v1/embeddings | Not a first-class route | No dedicated Copperway embeddings handler |
stream: true | Conditional | Disabled when PII masking is active; use stream: false |
OpenAI SDK | Drop-in | Set baseURL + apiKey (API secret) only |
Frontier APIs via gateway
Route Claude Fable 5, Opus 4.8, GPT-5.5, Gemini 3.1 Pro, xAI Grok 4.5, and partner models through a single OpenAI-compatible endpoint. PII is masked before any upstream call. Copperway is the live gateway; campus models are not shipping — Vasilikos is pre-construction.
- Claude Fable 5
- Claude Opus 4.8
- Claude Opus 4.8 Thinking
- GPT-5.5 xHigh
- GPT-5.5
- Claude Sonnet 5 High
Full catalog
Claude Fable 5 · Claude Opus 4.8 · Claude Opus 4.8 Thinking · GPT-5.5 xHigh · GPT-5.5 · Claude Sonnet 5 High · Gemini 3.1 Pro Preview · Claude Opus 4.6 · Claude Sonnet 4.6 · Claude Haiku 4.5 · GPT-4.1 · o3 · o4-mini · GPT-4o · Gemini 3 Pro · Gemini 2.5 Pro · Gemini 2.5 Flash · Grok 4.6 · Grok 4.5 · Mistral Large 3 · DeepSeek V4 Flash · Kimi K3 (max) · Qwen3.8 Max · GLM-5.2 (max) · MiniMax-M3
- Claude Fable 5
- Claude Opus 4.8
- Claude Opus 4.8 Thinking
- GPT-5.5 xHigh
- GPT-5.5
- Claude Sonnet 5 High
- Gemini 3.1 Pro Preview
- Claude Opus 4.6
- Claude Sonnet 4.6
- Claude Haiku 4.5
- GPT-4.1
- o3
- o4-mini
- GPT-4o
- Gemini 3 Pro
- Gemini 2.5 Pro
- Gemini 2.5 Flash
- Grok 4.6
- Grok 4.5
- Mistral Large 3
- DeepSeek V4 Flash
- Kimi K3 (max)
- Qwen3.8 Max
- GLM-5.2 (max)
- MiniMax-M3
Catalog · 15 Aug 2026 — synced from shared frontier-models data. Auto-refresh via GET /api/leaderboard/sync cron.
Arena quality scorecard
Arena ELO (reference) · 2026-07-28Top models from LMSYS Chatbot Arena — reference rankings only; AGICY does not run independent evals. Full leaderboard →
Pricing audit (sheet 05): Standard Gateway blended savings run 12.4–17.6% vs direct on flagship models at 75/25 mix; SRA intro is €0.10/M flat both directions. Negative Standard rows on budget SKUs reflect the €0.15/M platform floor — not a math error.
In-house design notes
Outcome notes from AGICY's own Copperway, planned Galaxy fleet, and EU compliance pathfinding. These are in-house build notes — not third-party client endorsements.
Green outlook
Semantic compression lowers billed token volume — and the energy/carbon footprint of sovereign inference on AGICY's Cyprus fleet. 46% fewer input tokens billed → 12.4–17.6% blended rate saving → up to ~54% total bill reduction. Live charts update as telemetry arrives.
Next: Specifications & FAQ →
Copperway FAQ
Common questions on the Copperway sovereign gateway — OpenAI compatibility, PII masking, semantic compression economics, and EU compliance — for engineering and procurement reviewers.
Copperway console
Configure upstream keys, test completions, and inspect live gateway health. Read-only endpoints are public; mutations require sign-in. Keys are encrypted at rest when PROXY_VAULT_KEY is set.
Ready for production sovereign inference?
Route through Copperway in minutes, or speak with our team about enterprise deployment on Cyprus sovereign compute.