NEAR AI has launched Gateway, a macOS application that connects developer tools including Claude Code, Codex and OpenCode to models available through NEAR AI Cloud. According to the official NEAR AI announcement, the app lets developers route coding sessions through different privacy tiers without manually reconfiguring each tool. Gateway is available now as a first, terminal-focused release aimed at development workflows involving proprietary code, unreleased products or other sensitive context.
The application itself runs on the user’s Mac, but that does not mean all AI inference occurs locally. Private models execute inside secure hardware enclaves, while Incognito models route requests to external AI providers using NEAR AI Cloud credentials. Incognito routing can reduce what the upstream provider knows about the user, but the external provider still processes the request itself. The distinction prevents Gateway’s privacy model from being treated as a single security guarantee covering every model in its catalog.
Gateway Adds Privacy Controls to Coding Workflows
Gateway launches terminal sessions with supported coding tools already configured to use NEAR AI Cloud. Developers can therefore continue working through familiar interfaces while selecting models according to the privacy requirements of a project. For supported Private models, prompts are processed inside Trusted Execution Environments rather than being exposed in plaintext to the surrounding cloud infrastructure under NEAR AI’s security model. This builds on the confidential-computing architecture already available through NEAR AI’s earlier OpenCode private inference integration.
Trusted Execution Environments do not mean that data is never decrypted. Prompts must become readable inside the protected enclave for inference, while the surrounding host environment is designed to remain unable to access that plaintext. Hardware attestation can provide evidence about the environment handling the workload, but it does not prove that the model output itself is correct. Confidential computing therefore relocates the trust boundary into protected hardware rather than eliminating every infrastructure dependency. NEAR AI has previously expanded that model through its verifiable confidential inference stack.
Users Control Which Session Traces Leave Their Mac
Gateway also introduces a separate data-sharing layer built around session traces. Users can inspect locally generated records showing information such as model usage, token consumption, repeatedly accessed files and session costs. Nothing from those traces is shared through the contribution program unless the user opts in, with redaction performed locally and each trace requiring approval before transmission. Accepted contributions can earn NEAR AI Cloud credits that users can apply toward future model inference.
That incentive system should be distinguished from the privacy of the inference session itself. A developer can use confidential inference for sensitive work and choose not to contribute the resulting traces, while sharing data from a separate open-source project. Gateway separates model execution from voluntary data contribution rather than making data sharing a condition of using the application. NEAR AI says future versions may add more applications beyond terminal tools, referral rewards, additional incentives for useful contributed data and more detailed personal session analysis.
The launch extends a wider NEAR AI strategy that already includes confidential coding, protected agent execution and privacy-focused financial infrastructure. NEAR has also introduced confidential payment infrastructure for AI agents, although those payment systems are separate from Gateway’s developer tooling. The measurable development today is narrower: macOS users can now access NEAR AI’s different inference privacy tiers directly through three established coding interfaces while retaining local control over whether their session traces are contributed. Adoption, model availability, latency and developer willingness to route sensitive production work through TEE-based infrastructure will determine how significant that integration becomes.
Natalie Pierce tracks the parts of crypto that move fast and rarely wait for everyone to catch up. From South Africa, she covers DeFi, AI crypto, hacks, airdrops, sentiment and emerging narratives, especially when user behavior and protocol risk start to overlap.
Her reporting is built for messy sectors. Natalie looks at incentives, reactions, security concerns, social momentum and early signs of traction without pretending every new trend is already proven. Her voice is clear and accessible, but careful enough for areas where excitement can outrun the facts very quickly.
