Skip to content
STIRUP

TECHNOLOGY INTELLIGENCE

Understand what's next.

New Open-Source AI Gateway for Self-Hosted AI Management

The New API is an open-source AI gateway designed as a self-hosted solution for managing applications, agents, and teams with features like authentication, analytics, and private deployment.

Share
AI-generated conceptual illustration accompanying: New Open-Source AI Gateway for Self-Hosted AI Management
AI-generated editorial illustration of the article topic; not a photograph, product screenshot or measured result.

Overview and Purpose of New API

The New API is an open-source AI gateway designed to serve as a self-hosted solution for managing applications, agents, and teams. Its primary purpose is to act as a centralized interface for AI models, enabling features such as organization-level authentication, multi-model management, usage analytics, and cost accounting. The project emphasizes private deployment, making it suitable for scenarios requiring secure and controlled access to AI capabilities. It is important to note that the project is explicitly intended for lawful and authorized use only, as stated by the developers. Additionally, WebSocket support is available but requires a compatible upstream and channel configuration, as noted in the documentation. This project provides a foundation for organizations seeking to deploy and manage AI services internally without relying on third-party platforms.

Deployment and Setup Instructions

To deploy the LACERBA/new-api-20261117 project locally, users can utilize Docker with the following command: mkdir -p data followed by docker run --name new-api -d --restart unless-stopped -p 127.0.0.1:3000:3000 -e TZ=Asia/Shanghai -v "$(pwd)/data:/data" calciumion/new-api:latest. This sets up a single instance with SQLite and binds it to localhost, as described in the project's Quick start section.

After deployment, users should navigate to http://localhost:3000 to complete the setup wizard and create an administrator account. For OpenAI-compatible clients, the base URL should be set to http://localhost:3000/v1, and the New API-issued key should be used for authentication, as outlined in the project's documentation.

To list accessible models, users can execute the following command: curl --fail-with-body http://localhost:3000/v1/models -H "Authorization: Bearer ${NEW_API_KEY}". To call responses, replace your-enabled-model with an enabled model that supports the interface, as noted in the project's Deployment section.

Technical Architecture and Components

The LACERBA/new-api-20261117 project is structured around a modular architecture designed for extensibility and separation of concerns. The router/, middleware/, and controller/ directories handle HTTP routing, access control, and API request handling. The relay/ directory manages upstream adapters and request routing, while service/ and model/ contain business logic and data persistence components. The relaykit/ module is a standalone Go library for protocol data transfer objects (DTOs) and conversions, enabling independent development and reuse. JavaScript task plugins are located in plugins/tasks/, with documentation available in the Task Plugin API v1. The web console is built using React 19, TypeScript, Rsbuild, TanStack, and Tailwind CSS 4, and the desktop wrapper is implemented in Electron. Contributors are advised to review AGENTS.md for guidelines.

The LACERBA/new-api-20261117 project is distributed under the GNU Affero General Public License v3.0 (AGPLv3), as stated in the project's license documentation. This license requires that any modified versions of the software preserve the author attribution notice, which credits the frontend design and development to New API contributors. Additionally, if a modified version includes a user interface, it must maintain a visible link to the original project at https://github.com/QuantumNous/new-api. These requirements ensure that users and developers remain connected to the project's origins and maintain proper attribution. The licensing terms are consistent with the broader open-source community's emphasis on transparency and collaboration. Developers using or modifying this project should review the full license text for detailed obligations and rights.

Comparison with Other Open-Source Gateways

In comparing LACERBA/new-api-20261117 with other open-source AI gateways, Bifrost emerges as a strong contender. According to getmaxim.ai, five gateways—Bifrost, LiteLLM, Kong AI Gateway, Apache APISIX, and Agent Router—are evaluated. Bifrost, maintained on GitHub under the Apache 2.0 license and developed by Maxim AI, is highlighted as the best choice for enterprises requiring high performance, scalability, and reliability. As noted in getmaxim.ai, Bifrost leads in raw performance, MCP support, and governance depth. Built in Go, the Bifrost AI gateway is designed for production-scale self-hosted deployments, as stated in getmaxim.ai. These attributes position Bifrost as a robust alternative to LACERBA/new-api-20261117 for organizations prioritizing enterprise-grade AI infrastructure.

Sources and reporting notes

Based on published source material. No independent hands-on testing is claimed.

Read more