One API for all AI models. Governed.
PII redaction. Zero prompt-content retention by default. EU data-residency tiers applied per request. The audit trail your DPO can hand to an AI Act auditor.
Dashboard
EU residency per request · 0 policy refusals| Time | Event | Actor | Model | Region | Detail |
|---|---|---|---|---|---|
| 14:03:21 | Allowed | jonatan.berg@acme.eu | claude-opus-4-7 | fr-par-1 | 1,247 in / 512 out |
| 14:03:20 | PII redacted | jonatan.berg@acme.eu | mistral-large | fr-par-1 | 1 email redacted |
| 14:03:19 | Denied | agent_3vQk_summarizer | gpt-4o | — | model host outside EU |
| 14:03:18 | Allowed | agent_bot_customer_support | mistral-devstral-sweden | se-sto-1 | routed via policy |
| 14:03:17 | Trace stored | anna.lind@acme.eu | claude-haiku-4-5 | fr-par-1 | trace t_30vPb · €0.0214 |
The problem with direct LLM access
Data exposure
Sending sensitive data to LLMs without proper controls creates GDPR risk. Without governance, organizations lack visibility into what data is being shared with which providers.
No audit trail
EU AI Act (Aug 2026) introduces record-keeping requirements for AI systems. Many organizations lack structured logging of AI usage — who asked what, when, and which model answered.
No territorial control
Most LLM providers process data in the US by default. Enforcing data sovereignty per request, per team, or per use case typically requires custom infrastructure.
Three capabilities. One endpoint.
Intelligent Gateway
Single API endpoint for all LLM providers. Smart routing by cost, speed, quality, and territorial policy. Supports text, image, and PDF input across every compliance tier. OpenAI-compatible format — no SDK changes required.
Model health
12 models · 30d| Model | Status | P50 | Success | Requests | Spend |
|---|---|---|---|---|---|
| claude-opus-4-7 | 187 ms | 100.0% | 4,218 | €812.40 | |
| mistral-large | 242 ms | 99.7% | 3,892 | €586.21 | |
| claude-sonnet-4-6 | 204 ms | 99.9% | 3,107 | €498.62 | |
| claude-haiku-4-5 | 112 ms | 100.0% | 2,104 | €201.14 | |
| mistral-devstral-sweden | 338 ms | 99.0% | 1,401 | €331.07 | |
| gemini-pro | 268 ms | 99.6% | 1,188 | €247.55 | |
| claude-opus-4-8 | 196 ms | 100.0% | 972 | €384.10 | |
| magistral-medium | 311 ms | 99.4% | 844 | €166.92 | |
| mistral-codestral | 159 ms | 99.8% | 703 | €97.41 | |
| mistral-medium | 984 ms | 98.4% | 621 | €128.00 | |
| gpt-4o | 221 ms | 99.5% | 402 | €118.73 | |
| nova-lite-sweden | 297 ms | 100.0% | 211 | €82.00 |
Per-Request Audit Trail
Requests logged with actor identity, model attribution, territorial proof, and cost. Searchable, exportable, designed to support compliance workflows.
Audit log
12,447 events · 30d · exportable| Time | Event | Actor | Model | Region | Detail |
|---|---|---|---|---|---|
| 14:03:21 | Allowed | jonatan.berg@acme.eu | claude-opus-4-7 | fr-par-1 | 1,247 in / 512 out |
| 14:03:20 | PII redacted | jonatan.berg@acme.eu | mistral-large | fr-par-1 | 1 email redacted |
| 14:03:20 | Allowed | anna.lind@acme.eu | claude-haiku-4-5 | fr-par-1 | routed via policy |
| 14:03:19 | Denied | agent_3vQk_summarizer | gpt-4o | — | model host outside EU |
| 14:03:18 | Allowed | agent_bot_customer_support | mistral-devstral-sweden | se-sto-1 | routed via policy |
| 14:03:17 | Trace stored | anna.lind@acme.eu | claude-haiku-4-5 | fr-par-1 | trace t_30vPb · €0.0214 |
| 14:03:16 | Allowed | jonatan.berg@acme.eu | mistral-large | fr-par-1 | 894 in / 301 out |
| 14:03:15 | Allowed | agent_bot_customer_support | claude-haiku-4-5 | fr-par-1 | routed via policy |
| 14:03:14 | Trace stored | anna.lind@acme.eu | mistral-devstral-sweden | se-sto-1 | trace t_30vQa · €0.0089 |
| 14:03:12 | Allowed | agent_3vQk_summarizer | claude-opus-4-7 | fr-par-1 | 2,031 in / 740 out |
EU Territorial Compliance
Routing tiers: EU-Sweden (single-country Stockholm residency), EU-Strict (EU-owned providers), EU-Cloud (EU infrastructure), Unrestricted. Enforced per request as hard constraints, configurable per team.
Projects
4 active · €2,140.82 total| Project | Status | Tier | Requests | Spend | Top model |
|---|---|---|---|---|---|
| Signal Hub | EU-Cloud | 5,621 | €1,084.20 | claude-opus-4-7 | |
| Lighthouse | EU-Cloud | 3,108 | €527.36 | mistral-large | |
| Atelier | EU-Sweden | 2,807 | €389.14 | mistral-devstral-sweden | |
| Outpost | EU-Cloud | 911 | €140.12 | claude-haiku-4-5 |
Everything between your application and the model layer.
One inline gateway resolves the right model, enforces your policy, emits an immutable audit record, and forwards the request to a provider on infrastructure that satisfies your contract.
One request, routed by your cost, speed and quality priorities.
Routes each request to a suitable model by your cost, speed and quality preferences, while hard-enforcing your residency tier.
- Cost-, latency- and availability-weighted selection — within your compliance tier.
- Per-team default tiers, per-request overrides
- Text, image and PDF input across every compliance tier
- Hard territorial constraints enforced before egress
Your app
Providers (EU)
claude-sonnet-4.6 via Bedrock eu-central-1 — chosen for capability within the eu_cloud constraint.Every call is signed evidence by the time it returns.
Requests logged with actor identity, model attribution, territorial proof, and cost. Searchable, exportable, designed to support compliance workflows.
- Actor identity per request (customer-asserted)
- Model, region and territorial tier recorded per call
- Token counts and EUR cost on every span
- Searchable, exportable audit log
Two lines of code. Your stack doesn’t notice.
Change api.openai.com to api.brainorchestra.ai and add an actor token. Your existing OpenAI SDK code keeps working — requests are tracked automatically.
- Drop-in OpenAI Chat Completions schema
- Python, TypeScript, Java, Go, Ruby — same patch
- Bring your own provider keys, or use ours
- Audit log fills within seconds of first call
# your existing openai client — change two lines
from openai import OpenAI
client = OpenAI(
api_key="bo_live_…",
base_url="https://api.brainorchestra.ai/v1",
default_headers={"X-Actor-Token": "bo_actor_…"},
)
resp = client.chat.completions.create(
model="eu-strict", # or a specific id
messages=[{"role": "user", "content": "Summarize this contract"}],
)Integrate in 5 minutes
Sign up
Create your account and set your territorial policy. Choose EU-Strict, EU-Cloud, or Unrestricted as your default.
Swap your API URL
Change api.openai.com to api.brainorchestra.ai
Requests are tracked
Audit trails, territorial routing, cost tracking — automatic per request through the gateway.
from openai import OpenAI
client = OpenAI(
api_key="$BO_API_KEY",
base_url="https://api.brainorchestra.ai/v1",
default_headers={"X-Actor-Token": "bo_actor_YOUR_ACTOR_TOKEN"},
)One endpoint for the whole catalog.
Chat models from Mistral, Anthropic, OpenAI, Amazon, Google and more — every provider reachable through the same OpenAI-compatible endpoint, each routed within the tiers it can serve.
Mistral
Evroc
Amazon
GoogleCompliance routing built in
EU-Strict
Models served by EU-owned providers on EU infrastructure. Open-weight models run locally — no service is procured from the model's author.
EU-Cloud
Claude + Mistral + Cohere + Amazon Nova on EU infrastructure via AWS Bedrock (Frankfurt or Stockholm).
Unrestricted
All models, all providers. Maximum capability.
Routing shorthand
"model": "eu-strict" # Best EU-owned model
"model": "eu-strict:small" # Cheapest EU model
"model": "auto" # Best for your tier
"model": "mistral-large" # Specific modelBrain Orchestra provides technical controls and structured evidence. It does not grant compliance — your legal team makes that determination.
Simple, transparent pricing
Prepaid credit balance with transparent per-request billing. Bring your own provider keys, or use managed keys.
- All features unlocked
- Unlimited projects and users
- BYOK: no charge at all
- Managed keys: billed from your prepaid balance
- Card required at signup
- Everything in Trial
- 100,000 requests / month
- All EU-sovereignty tiers (eu_strict, eu_cloud, eu_sweden)
- Mixed key model (BYOK + BO-managed)
- Managed keys: billed from your prepaid balance
- BYOK: zero per-token fee
- Everything in Pro
- Unlimited requests
- Custom convenience fee
- SLA + dedicated support
- SSO/SAML (roadmap — ask sales)
- VPC / on-premise (roadmap — ask sales)
Also available in USD ($79/mo) and SEK (899 kr/mo). Currency selected at signup. All prices exclude VAT where applicable.
Built for agentic AI
When AI agents call AI agents, you need governance at every step. Brain Orchestra’s trace system gives you operation-level budgets, depth limits, kill switches, and a full span tree for every agent workflow.
- BudgetPer-operation spend limits with automatic cutoff
- Depth & fan-outCap agent-to-agent nesting and parallel calls
- Kill switchTerminate any running workflow instantly