Designed to reduce unnecessary context without degrading the model response — and measured per tool-set, per request. OpenAI-compatible shapes, published rates, and the catalog in Studio — chat, voice, and code on one key.

The Free plan is $0. Notre Runtime is included on every plan.

+ New conversation

Today

MSA renewal dates
RAG chunking strategy
n8n usage guard

Yesterday

Voice prototype script
Embed migration plan

anthropic/claude-haiku-4.5

Weekly credits · 4,486 left

Summarize this 40-page MSA and list every renewal date.

caedral · anthropic/claude-haiku-4.5

Three auto-renewals: the hosting amendment (Mar 1), the data processing addendum (Apr 15), and the support schedule (Jun 30). Everything else expires without renewal.

Drafting the renewal calendar now▍

Notre Runtime·12,847 tokens saved

Ask anything, or / for commands

One key, three products

Everything Caedral builds, on one key.

The laboratory where you try models, the compiler that keeps chat cheap, and the OpenAI-compatible surface you ship on.

The AI laboratory
Caedral Studio
Chat, voice, images, video, and code in the browser — metered in weekly plan credits as you use it.
See pricing
Economy on chat
Notre Runtime
Up to 99.7% fewer tokens in validated Notre V4 MAX workloads — with required-set recall held at 100%.
How the compiler works
OpenAI-compatible
API
One key, one request shape. POST /v1/chat/completions on prepaid at-cost rates with the Notre Fee — never a surprise.
Read the quickstart

The API, live

The exact requests your app sends — switch product and language in the demo. One base URL, OpenAI-compatible shapes, published rates.

Read the quickstart

POST/v1/chat/completions

Request

curl https://api.caedral.com/v1/chat/completions \
  -H "Authorization: Bearer $CAEDRAL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "anthropic/claude-haiku-4.5",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

import { Caedral } from "caedral"

const client = new Caedral({
  apiKey: process.env.CAEDRAL_API_KEY,
})

const completion = await client.chat.completions.create({
  model: "anthropic/claude-haiku-4.5",
  messages: [{ role: "user", content: "Hello" }],
})

from caedral import Caedral

with Caedral() as client:
    completion = client.chat.completions.create(
        model="anthropic/claude-haiku-4.5",
        messages=[{"role": "user", "content": "Hello"}],
    )

Result · illustrative Notre workload

Status: 200 OK
Model: anthropic/claude-haiku-4.5
Input sent: 5,103
Output: 42
Notre: 7,739 saved

Notre Runtime applies on chat completions only. Embed, Rerank, and Voice are different request shapes.

POST/v1/embeddings

Request

curl https://api.caedral.com/v1/embeddings \
  -H "Authorization: Bearer $CAEDRAL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "caedral-embed-e1-small-v1",
    "input": "Retrieval-ready vectors"
  }'

import { Caedral } from "caedral"

const client = new Caedral({
  apiKey: process.env.CAEDRAL_API_KEY,
})

const result = await client.embeddings.create({
  model: "caedral-embed-e1-small-v1",
  input: "Retrieval-ready vectors",
})

from caedral import Caedral

with Caedral() as client:
    result = client.embeddings.create(
        model="caedral-embed-e1-small-v1",
        input="Retrieval-ready vectors",
    )

Result

Status: 200 OK
Model: caedral-embed-e1-small-v1
Dimensions: 384

Notre Runtime does not compile this endpoint.

POST/v1/rerank

Request

curl https://api.caedral.com/v1/rerank \
  -H "Authorization: Bearer $CAEDRAL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "BAAI/bge-reranker-v2-m3",
    "query": "Caedral billing",
    "documents": ["Weekly plan credits", "Prepaid balance"]
  }'

import { Caedral } from "caedral"

const client = new Caedral({
  apiKey: process.env.CAEDRAL_API_KEY,
})

const ranked = await client.rerank.create({
  model: "BAAI/bge-reranker-v2-m3",
  query: "Caedral billing",
  documents: ["Weekly plan credits", "Prepaid balance"],
})

from caedral import Caedral

with Caedral() as client:
    ranked = client.rerank.create(
        model="BAAI/bge-reranker-v2-m3",
        query="Caedral billing",
        documents=["Weekly plan credits", "Prepaid balance"],
    )

Result

Status: 200 OK
Model: BAAI/bge-reranker-v2-m3
Top result: index 0 · 0.91

Notre Runtime does not compile this endpoint.

POST/v1/audio/speech

Request

curl https://api.caedral.com/v1/audio/speech \
  -H "Authorization: Bearer $CAEDRAL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "caedral-voice-1",
    "input": "Caedral Voice",
    "voice": "caedral-f1"
  }'

import { Caedral } from "caedral"

const client = new Caedral({
  apiKey: process.env.CAEDRAL_API_KEY,
})

const speech = await client.audio.generate({
  model: "caedral-voice-1",
  input: "Caedral Voice",
  voice: "caedral-f1",
})

from caedral import Caedral

with Caedral() as client:
    speech = client.audio.generate(
        model="caedral-voice-1",
        input="Caedral Voice",
        voice="caedral-f1",
    )

Result

Status: 200 OK
Model: caedral-voice-1
Body: audio/wav

Notre Runtime does not compile this endpoint.

Caedral Studio

The laboratory is opening soon.

See pricing

Seven labs in the browser — chat, voice, images, videos, code, sheets, and the showcase. Every request is metered to the same weekly credits and prepaid balance as the API, with no separate key.

anthropic/claude-haiku-4.5

Pick the cheapest model that holds 128k context.

Two qualify under 128k — here is the rate comparison▍

Notre · 6,412 tok saved

Chat

Every model in one thread. Notre trimming and credit debits are shown per turn.

See pricing

0:09 · WAV

Voice

Text-to-speech and transcription against Caedral Voice — 4 house voices, metered to your subscription.

See pricing

landing-page/

index.html

app.js

styles.css

// generated from one prompt · Code Lab

export async function ask(prompt) {

const r = await fetch("/v1/chat/completions", {

method: "POST", body: JSON.stringify({ model, messages }),

});

return r.json();

}

A pricing page with one amber CTA…

Generate

Code lab

Prompt-to-code sessions with a live file tree and editor — no extra key, no separate billing surface.

See pricing

Images & videoText-to-image and video generation on the same key.

Voice ArsenalAudition the 4 house voices — no account needed.

Built by Caedral

Our products and the frontier, classified.

Frontier labs are in the catalog. These four are developed or hosted by Caedral. Classification is explicit — we do not hide our own work behind someone else's name.

Caedral Embed

caedral-embed-e1-small-v1

384-dimensional embeddings. Developed and operated by Caedral. POST /v1/embeddings.

Plan credits

Caedral Rerank

BAAI/bge-reranker-v2-m3

Hosted by Caedral over an external foundation model. POST /v1/rerank.

Prepaid

Caedral Voice

caedral-voice-1

English WAV speech. Developed and operated by Caedral. Notre Runtime does not apply. POST /v1/audio/speech.

Plan credits

Notre Runtime

eligible chat

Native context compiler on POST /v1/chat/completions. Not a SKU. MCP is an integration, not the product name.

Included

Catalog

Live model IDs. Published rates.

Full catalog

682 models across 94 providers, plus the Caedral-operated products above.

Model	ID	Context
Anthropic: Claude Fable Latest	~anthropic/claude-fable-latest	1,000,000
Anthropic: Claude Haiku Latest	~anthropic/claude-haiku-latest	1,000,000
Anthropic: Claude Opus Latest	~anthropic/claude-opus-latest	1,000,000
Anthropic: Claude Sonnet Latest	~anthropic/claude-sonnet-latest	1,000,000
DeepSeek: DeepSeek Flash Latest	~deepseek/deepseek-flash-latest	1,048,576
DeepSeek: DeepSeek Pro Latest	~deepseek/deepseek-pro-latest	1,048,576
DeepSeek: DeepSeek V4 Flash Latest	~deepseek/deepseek-v4-flash-latest	1,048,576

Notre Runtime

Measured. Not promised.

Designed to reduce unnecessary context without degrading the model response — and measured per tool-set, per request. Up to 99.7% fewer tokens in validated Notre V4 MAX workloads — with required-set recall held at 100%.

Fewer input tokens99.7%

M1.4 V4 MAX corpus — the honest bound, including a schema round-trip per required tool.

Blind-holdout recall99.47%

After the policy freeze (564/570) — the gate any stronger claim must pass.

Frozen safety regression192/192

Held — with required-set recall at 100% on the development corpus.

Verified benchmark — M1.4 development corpus · V4 MAX — Notre Runtime

01Large request context
02Notre Runtime
03Reduced model input
04Model inference
05Result

Input before: 567,755,348
Input sent: 1,894,662
Saved: 565,860,686

99.7%reduction

TRUE AGGREGATE on the recorded corpus (n=4,962). Output tokens are generated by the model and are not claimed as Notre savings.

Notre Runtime is built into Caedral. No add-on. No separate bill. It is applied on eligible POST /v1/chat/completions when the platform runs Notre — never on embeddings, rerank, voice, or images. Different request shapes do not share this compiler, and we do not pretend they do.

The Notre program Notre Runtime Full results

Notre economics

Scale the verified ratio yourself.

Inspect the recorded compiler benchmarks, then project the same ratio onto a realistic input budget at Caedral prices. The model control sets the current tariff — Notre was not A/B tested on every catalog model.

Verified Notre benchmark

Same plan. Same price. More useful AI work.

Notre Runtime is built into Caedral. No add-on. No separate bill. Reduction is from the compiler corpus — not a live A/B of the catalog model you pick. Dollars use that model's current Caedral input tariff.

Up to: 99.7%
Median: 81.4%
Notre Runtime: Included

Benchmark

Model (Caedral input tariff)

Economic scale

Projected from verified benchmark reduction. Same ratio, normalized workload.

Same budget

299.66×

equivalent input workload

01Input before Notre
02Notre Runtime compiles
03Input sent to the model

Input before10,000,000
Input sent33,372
Input saved9,966,628

Reduction: 99.7%
Quality: Required-set recall 100%

Projected from verified benchmark

Input tokens avoided

9,966,628

of 10,000,000 baseline input

Usage value preserved · Anthropic: Claude Sonnet 4.6

$29.90

Current Caedral input tariff only. Output is not counted as savings.

More use from the same plan

100→29,966

equivalent requests on the same input budget

Compiler time (not model round-trip)

p50 10.5 ms · p95 50.7 ms

$10 prepaid budget

Without Notre-equivalent: 3,333,333 input tokens.

With this reduction: 998,868,208 equivalent input tokens.

Same budget, projected from the verified ratio and this model's current input price.

At this benchmark's 99.7% reduction rate, a 10,000,000-token input allowance can process roughly 299.66× the equivalent workload before exhausting the same input budget.

Benchmark details

Workload: M1.4 development corpus · V4 MAX
Date: 2026-10-03
Runs: 4,962
Policy: V4 MAX
Input before (recorded): 567,755,348
Input after (recorded): 1,894,662
Quality methodology: 4,962 of 4,962 required tool-sets preserved (0 failures; Wilson 95% CI [99.92%, 100%]). Reduction is the honest primary metric: level-1 disclosure payload plus oracle round-trips for every required tool (1.08 round-trips/case). Pessimistic bound (round-trip for every selected tool): 98.51%. Deterministic lexical router — no LLM in the routing path.
Source: caedral-notre-mcp/benchmarks/results/M1.4-GUARDED-V3.json#v4_max_candidate.m14_development
Confidence: TRUE AGGREGATE SUM/SUM. Blind holdout pending single post-freeze evaluation. Development corpus tuned the lexical triggers; production shadow validation required before default-on.
Fallback rate: 0.0%

Full verified table

Live on the platform

Models in the catalog: 682
Providers behind one key: 94
House voices in the Arsenal: 4
Caedral-operated products: 4
Verified Notre input reduction: 99.7%

Get started

Start with one API key

Create an account, pick a plan, and route chat, embeddings, rerank, and Notre-enabled completions through Caedral. Notre Runtime is built into Caedral. No add-on. No separate bill.

Get API key Read quickstart

The lab. The economy. The API. One key.

Caedral Studio

Notre Runtime

API

Chat

Voice

Code lab

Measured. Not promised.

Start with one API key