# System One API

> Hosted API returning structured choices, scores and probabilities for AI agent decisions

- Category: [Developer Tools](https://finderatlas.com/categories/developer-tools)
- Pricing: Free trial
- Website: https://system-one.dev/en/api
- Platforms: API, Web
- API: available
- Last updated: 2026-10-25
- FinderAtlas page: https://finderatlas.com/tools/system-one-api

## Overview

System One is a hosted API and open-source TypeScript SDK for bounded decisions inside AI agents and applications. A single endpoint (POST /v1/systemone) takes a shared "state" such as a customer message, plus up to 32 questions, and returns structured answers: a selected option with a probability for each option (Choice), a score on a 2-10 level rubric (Score), or a yes/no probability (Noul). Your application decides what to do with the result; the service does not execute tools.

## Key features

- Choice
- Score
- Noul
- Multi-question requests
- Model catalogue
- TypeScript SDK
- Idempotency and error handling
- Credit billing

## Pricing

- **Starter** ($5.99/one-time): 5,000 prepaid decision credits.
- **Builder** ($19.99/one-time): 20,000 prepaid decision credits.
- **Scale** ($99.99/one-time): 120,000 prepaid decision credits.

## Best for

- Routing and triage of requests
- Ranking candidates against a rubric
- Gating agent actions with yes/no checks

## Who uses it

- Developers
- AI agent builders
- Backend engineers

## Strengths

- Pay-as-you-go credits with no subscription and no expiry
- Failed calls are not charged
- Public model catalogue with prices and limits
- Structured outputs with per-option probabilities
- Open-source MIT-licensed TypeScript SDK

## Limitations

- Early-stage project: the SDK repository has very few GitHub stars and the company is not identified on the site
- No published latency, uptime or accuracy guarantees
- Limited to classification, scoring and yes/no style decisions rather than generation
- Hosted models are the vendor's own and independent benchmarks are not provided

## Not recommended for

- Open-ended text generation
- Workloads needing published uptime or latency guarantees
- Non-technical users

## FinderAtlas editorial scores (out of 10)

- API Design: 8/10
- Documentation Clarity: 7.8/10
- Pricing Transparency: 8.5/10
- Maturity & Reliability: 5.8/10
- Use Case Breadth: 6.2/10

## Alternatives

- [Browserbase](https://finderatlas.com/tools/browserbase): Cloud browser infrastructure and APIs that let AI agents browse and act on the web ([compare](https://finderatlas.com/compare/browserbase-vs-system-one-api))
- [Browser Use](https://finderatlas.com/tools/browser-use): Open-source library and cloud platform for AI agents that control real web browsers ([compare](https://finderatlas.com/compare/browser-use-vs-system-one-api))
- [Genspark](https://finderatlas.com/tools/genspark): AI super agent workspace that turns one prompt into slides, docs, sheets and research ([compare](https://finderatlas.com/compare/genspark-vs-system-one-api))
- [Manus](https://finderatlas.com/tools/manus): General AI agent that plans, researches, codes and delivers finished work autonomously ([compare](https://finderatlas.com/compare/manus-vs-system-one-api))
- [Lindy](https://finderatlas.com/tools/lindy): AI teammate that handles email, meetings and recurring work across your team's tools ([compare](https://finderatlas.com/compare/lindy-vs-system-one-api))
- [Gumloop](https://finderatlas.com/tools/gumloop): Build AI agents that work across your company's apps, with IT in control ([compare](https://finderatlas.com/compare/gumloop-vs-system-one-api))
- [CrewAI](https://finderatlas.com/tools/crewai): Open-source multi-agent framework and enterprise platform for building and running AI agent workflows ([compare](https://finderatlas.com/compare/crewai-vs-system-one-api))
- [n8n](https://finderatlas.com/tools/n8n): Build and self-host workflow automations and AI agents with visual nodes and custom code ([compare](https://finderatlas.com/compare/n8n-vs-system-one-api))

## Recent updates

- 2026-09-18: The public model catalogue lists the Jev model with this release date

## FAQ

**What does the API return?**

Structured answers for each question: the selected option with per-option probabilities for Choice, a level for Score, and a probability for Noul.

**Does it take actions for me?**

No. The site states that model output informs a decision and does not execute a tool or grant permission; your application controls every action.

**How is it priced?**

Prepaid credits, from $5.99 for 5,000 credits. Cost is questions times 4KB blocks times the model's rate. There is no subscription and credits never expire.

**Is there a free allowance?**

New accounts get 100 free credits on sign-up, according to the pricing page.

**Is the SDK open source?**

Yes. @system-one-ai/core is published on npm under the MIT license, with source on GitHub.

**Are there uptime or accuracy guarantees?**

No. The About and terms pages say no latency, uptime or individual prediction guarantee is published, and recommend evaluating accuracy on your own data.