Build in a weekend.
Scale to millions.

Start with a prompt. Add models, knowledge, tools, evaluations, security, and observability. Amarsia gives you the production infrastructure to move from prototype to reliable AI product.

01Build02Connect03Evaluate04Deploy05Monitor

Assisted prompt engineering

Design structured instructions, variables, reasoning, and output schemas with AI assistance—then keep full control over every change.

SYSTEM INSTRUCTIONS

Create new promptImprove prompt

Version Control & Rollback

Publish confidently. Every prompt and configuration update is traceable and reversible.

v18ProductionNow

v17Previous2h

v16Archived1d

Any Provider

Choose the right model per assistant. Switch providers without rebuilding your integration.

+ More

Bring your own keys

Knowledge Base

Turn documents and websites into reliable context for AI. Our vision enabled document processers makes each document into LLM ready knowledge instantly usable by its agents.

Supported sources

PDFDOCXPPTXXLSXCSVTXTWebsites+ More

Domain-specific documents are transformed and optimized into a structure best for LLM. We've data processing algorithms tuned for:

General PurposeLegalFinanceAccountingMedical

AI Actions

Build executable tools as actions for your agent. Start with our integrations and templates, or connect any MCP server.

REST API

GraphQL

+ Many more

Real-time web searchKnowledge base search

Client tools

Define and pass any client capability as an available action to the model at runtime.

Evaluate every run

Catch hallucinations and quality failures. Get logs and alerts for every model call against your evaluation metrics.

RUN #11794/4 passed

Answer relevancePass

GroundednessPass

Tool accuracyPass

Test cases

Ship every iteration safely with predefined test cases. Bulk run tests to pressure-test your agent.

Test suite▶ Run all

Refund policyPass

Support optionsPass

Operate with context

Inspect runs, latency, evaluation results, and action logs in one timeline. Iterate with confidence backed by data.

LOG #1179PASS

Model

Gemini 2.5

Latency

3.62s

Evaluations

4/4

Trigger agent

Trigger personalized agents via code from your backend or a secured link created in the dashboard.

Enterprise security

Your AI. Your data. Your control.

Protect prompts, responses, and enterprise data with encrypted infrastructure, deployment-level authentication, and domain allowlists for browser experiences.

Encryption

TLS encryption for data in transit
AES-256 encryption at rest
Customer-managed encryption keys
Private cloud storage

Data privacy

Your prompts and business data remain your IP
Least-privilege access controls
Secrets stored in a dedicated manager
No model training on customer data

Secure orchestration

Encrypted requests across every network hop
Processing inside isolated cloud compute
No persistent logging of decrypted content
Dedicated deployments available

Compliance

Regional data residency
Data Processing Agreements
Rolling operational-log deletion
Compliance roadmap for enterprise teams

Isolated compute

Customer-managed keys

Regional residency

Deployment controls

Developer experience

Use Amarsia with
React.

React hooks for streaming, conversations, and client tools — or a typed JavaScript SDK for server applications.

Read the SDK quickstart

import { useAgent } from "@amarsia/react"
import client from "@/lib/amarsia"

const { agent, messages, status } = useAgent(client)

await agent.start({
  content: [{ type: "text", text: "Help this customer." }],
})

npm install @amarsia/react @amarsia/sdkTyped responsesStreaming ready

Join the community

Discover what builders have to say about their Amarsia experience.

Join us on Discord

Build in a weekend.Scale to millions.

Platform features