Casturi models are fast, sized to the task and private by design. Talk to them in Casturi Chat, let them write and deploy code from your terminal with Casturi Code, or build them into your own product with the Casturi API.
Casturi 2.0CODE · Developed by Team Casturi · v0.84.3
from openai import OpenAI
client = OpenAI(
base_url="https://apiv2.casturiai.com/v1",
api_key=CASTURI_API_KEY,
)
stream = client.chat.completions.create(
model="casturi2.0-fast",
messages=[{
"role": "user",
"content": "Triage: my refund still hasn't arrived",
}],
stream=True,
)An assistant for research, writing, analysis and agents, with web search and your files built in.
See Chat in actionA coding agent in your terminal that plans, builds and deploys apps and websites with you.
Watch a sessionOpenAI-compatible endpoints with streaming, tool calling and zero data retention.
Try a requestHomework hints, learning games and coding for ages 7–17, with controls for parents.
Meet their learning buddyA learning buddy that helps them think things through. Homework hints, stories, 20 learning games and Code Lab, in a space designed for kids and teens.
Explore Casturi KidsA parent sets up the account. An ongoing free plan is available.
“How much is ¾ of 12?”
Let’s find one quarter first. What happens if you split 12 into four equal groups?
Example learning conversation
Daily limits and bedtime, for each child.
Topics, helpers and learning activity.
Practice, questions and a chance to try again.
Ask anything, search the web, work through your files, run code and hand repeatable work to agents. These are real screens from Casturi Chat. Select the orange markers to see what each part does.
Casturi breaks hard topics into headings, steps and definitions, so a long answer reads like a good explainer instead of a wall of text. Switch models, branch a reply or have it read aloud without leaving the conversation.
5 things to explore in this screen
Formulas are typeset as proper notation, inline and in blocks, so science, finance and engineering answers stay readable. Processes come back as ordered steps you can act on.
3 things to explore in this screen
Start from agents built for HR, finance, research, IT and sales, or make your own in the agent builder with instructions, files and tools. No code required. Share the ones your team should use.
4 things to explore in this screen
Add a custom connector by its URL and choose how it signs in. Your agents can then look things up and take action in your own systems, and you confirm you trust a connector before it's added.
4 things to explore in this screen
Casturi Chat follows your system theme, or switch it from the sidebar. Drag the handle to compare the agent marketplace in both.
Describe the app, website or deployment you want. Casturi Code plans the work, writes the code and runs the commands, and you can stop it at any moment. Pick a command to watch an example session.
CODE · Developed by Team Casturi · v0.84.3
A real Casturi Code start screen. Select the markers to see where commands, shortcuts and the live status line live.
The Casturi API follows the OpenAI chat completions format, so the SDKs, frameworks and tools you already use keep working.
import os
from openai import OpenAI
client = OpenAI(
base_url="https://apiv2.casturiai.com/v1",
api_key=os.environ["CASTURI_API_KEY"],
)
stream = client.chat.completions.create(
model="casturi2.0-fast",
messages=[{"role": "user", "content": "Classify this ticket: refund not received"}],
stream=True,
)
for chunk in stream:
if chunk.choices:
print(chunk.choices[0].delta.content or "", end="")import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://apiv2.casturiai.com/v1",
apiKey: process.env.CASTURI_API_KEY,
});
const res = await client.chat.completions.create({
model: "casturi2.0-fast",
messages: [{ role: "user", content: "Classify this ticket: refund not received" }],
});
console.log(res.choices[0].message.content);curl https://apiv2.casturiai.com/v1/chat/completions \
-H "Authorization: Bearer $CASTURI_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "casturi2.0-fast",
"messages": [{"role": "user", "content": "Classify this ticket: refund not received"}]
}'The largest general models are built for the hardest open-ended problems, and priced for them. Most business work is extracting, classifying, summarizing, drafting and coding to a spec. That work needs speed, accuracy and consistency more than sheer scale, so Casturi builds models sized for it.
| Area | One giant model for everything | Casturi |
|---|---|---|
| Cost | Frontier pricing on every request, including the simple ones. | You pay for the capability the task actually needs. |
| Speed | Larger models take longer to respond. | Right-sized models answer quickly. |
| Your data | Often retained for a period, and sometimes used for training. | Zero retention on the API, never used for training. Beta Chat conversations may train the model. See our Privacy Policy. |
| Improvement | Large releases, months apart. | Continuous training, reviewed by a supervisor model. |
| Domain depth | A generalist across every field. | Domain models for fintech, enterprise, industry policy and pharma, in training now. |
Most models improve in large, infrequent jumps. Casturi trains continuously, and a supervisor model reviews every cycle, so quality builds round after round instead of drifting.
The Casturi model learns from curated training data prepared by Team Casturi.
A separate supervisor model reviews the model's answers, scores them and pinpoints where it's weak.
The supervisor's corrections become the next round's training signal, aimed at those weak spots.
A new version ships only after it beats the current one on our evaluations. Then the cycle starts again.
During the invited beta, your Casturi Chat conversations may be used to improve this cycle. Casturi API requests are never part of it. See our Privacy Policy.
Privacy isn't a setting you have to find. Here's what happens to a request you send to the Casturi API.
Encrypted in transit over TLS.
On infrastructure we operate, with no outside AI providers.
Streamed back as it's generated.
Not stored, reviewed or used for training.
During the invited beta, Casturi Chat conversations may be used to train and improve the model. The Casturi API is never used for training. See our Privacy Policy.
Conversations are saved only so you can come back to them. Delete one and it's gone, or use a temporary chat that stays out of your history.
Casturi models are developed by Team Casturi and run on infrastructure we operate. Your prompts aren't passed to third-party AI services.
Casturi models and their weights are proprietary and never leave our control.
TLS protects every connection, and stored credentials are encrypted at rest.
Accounts are invite-only during preview, and every administrative action is recorded in an audit log.
Team Casturi is training specialist models for four fields where precise terminology, accuracy and compliance matter most. Each follows the same continuous, supervised training.
For banks, lenders, payments and wealth teams.
For operations, legal, HR and support teams.
For compliance, risk and governance teams.
For regulatory, clinical and safety teams.
Casturi is in invited preview while we scale. Request access now to be among the first in.
Every account starts on Free. During the preview, upgrades are granted by Team Casturi with no payment step.
10
messages a day
Try Casturi Chat with every feature switched on.
Request an invite100
messages a day
For people who rely on Casturi every working day.
Ask for ProNo cap
on daily messages
For heavy daily use across research, writing and code.
Ask for UnlimitedCustom
tokens and deployment
API keys, a startup program with 100M free tokens, dedicated deployments and early access to domain models.
Talk to Team CasturiCasturi is developed by Team Casturi. We design, train and serve our own closed models, and build Casturi Chat, Casturi Code and the Casturi API around them.
A coding agent that runs in your terminal. Describe what you want and it plans the work, writes the code, runs commands and can deploy to AWS, Azure or Google Cloud with a plan you review first. Choose "Casturi Code" in the access form to try it.
For the Casturi API, your request and its response are processed to deliver the answer, then discarded. In Casturi Chat, your conversation history is saved only so you can return to it. Delete conversations whenever you like, or use temporary chats.
The Casturi API is never used to train Casturi or any other model. During the invited beta, Casturi Chat conversations may be used to train and improve the model. See our Privacy Policy for details.
Most everyday work, like extraction, classification, summaries, drafting and code, doesn't need frontier scale. A model sized and trained for that work answers faster, costs less and gives more consistent results.
Yes. Set the base URL to https://apiv2.casturiai.com/v1, use your Casturi API key and set the model to casturi2.0-fast.
Models for fintech, enterprise, industry policy and pharma are in training now and coming soon. Request access and choose "Enterprise and domain models" to hear first.
Casturi is in invited preview. Request access below and we'll email your invite. You can then sign up with your email or Google account.
Tell us how you'd like to use Casturi. Team Casturi reviews requests daily and emails your invite when access is ready.
Already have an account? Log in to Casturi Chat