Axior CTN
Axior CTN · Multi-LLM Gateway · public APIs

The Creation Layer for all AI
One endpoint.
All the models.
Specialists for every task.

Routes tasks to the best public LLM — Claude 3.5 for reasoning, Gemini 1.5 for vision/OCR, Cohere for RAG, Mistral for fast code — with safety, fallbacks, and an EU sovereignty switch.

Why CTN is different

Specialists per task. Results: +20–30% quality, −30–40% cost, −30–50% latency.

Specialists per task

Reasoning → Claude; Vision → Gemini; RAG → Cohere; Code → Mistral.

Quality +Latency −

Failovers & caps

If a vendor throttles, CTN falls back automatically. Daily token caps control spend.

ResilienceCost control

EU sovereignty

Flip a flag to run on EU/open-weight models.

GDPR-friendly

Try the API

OpenAI-compatible. Point your client to one URL.

curl -s -X POST https://<app>.onrender.com/v1/chat/completions \
  -H "Authorization: Bearer YOUR_MASTER_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"claude-3-5-sonnet-20240620","messages":[{"role":"user","content":"Summarise CTN routing in 3 bullets."}]}'

Routing at a glance

Reads the task → selects a model from ctn-routing.yaml.

Option 1 (public APIs)

  • Reasoning: Claude 3.5 Sonnet → GPT-4o → Mistral
  • RAG: Cohere Embed+Rerank → Command-R+ (fallback Claude/GPT-4o)
  • Vision/OCR: Gemini 1.5 Pro → GPT-4o-mini
  • Codegen: Mistral Medium → GPT-4o → Claude

Security & governance

  • Master API key + CORS allowlist
  • Pre-filter policy hook
  • Daily token caps

CTN modules

Core modules with ready prompts.

Guided Funnel

  • Intent → structured spec
  • Clarifiers only if needed
  • JSON output

DSL Translator

  • NL ⇄ DSL
  • Strict grammar
  • 2 unit tests

RAG Answerer

  • Citations required
  • No fabrication
  • Confidence

Agent Planner

  • ≤6 steps
  • Tools only if needed
  • Rollback path

Codegen

  • Typed code + tests
  • Unified diff
  • ENV guard

Ops & News

  • Ops: P95, errors, usage
  • News: 5 bullets, links
  • Why-it-matters

Pricing guidance

Pay vendors directly; route cheap tasks to low-cost models.

Starter

£0 gateway

Use your vendor keys

Pro

£49/mo

Routing & alerts

Enterprise

Custom

SLA + on-prem

FAQ

Do I need accounts with each LLM vendor?

Yes. CTN uses your keys for OpenAI, Anthropic, Google, Cohere, Mistral.

Will modules break if models change?

No. Swap models in routing without touching your module code.

Is there analytics?

Add Langfuse/Helicone for latency, tokens, and cost analytics.

EU data?

Set the sovereign switch to route to EU/open-weight models.