OneCLI
An open-source, self-hosted enterprise AI assistant agent framework focused on secure sandboxed execution. It provides three layers of security: sandbox isolation, policy control, and zero credential exposure. It supports Slack and web interfaces and targets enterprise-scale AI assistant deployment scenarios.
Core Facts
Timeline (last 90 days)
Agent frameworks that emphasize security, controllability, and self-hosting are likely to become an important component of enterprise AI infrastructure
Each AI agent in OneCLI runs in an isolated sandbox environment, with task execution confined to a controlled scope.
OneCLI provides a policy-based control system that allows enterprises to granularly manage agents' operational permissions, accessible resources, and approval actions.
OneCLI adopts a zero-credential-touch design, where agents never directly access real credentials, reducing the risk of credential leakage at the architectural level.
Sandbox isolation design can limit the impact scope when an agent encounters a malicious prompt injection attack
OneCLI's self-hosted mode allows enterprises to deploy the entire agent framework within their own infrastructure, keeping data within the internal network, and leveraging its open-source nature for code-level transparency and auditability
OneCLI targets medium-to-large teams with security and compliance requirements that wish to introduce AI assistants at scale
OneCLI received 86 upvotes after launching on Product Hunt, ranking 13th for the day.
OneCLI was categorized on Product Hunt under tracks including open source, artificial intelligence, GitHub, and virtual assistants.
OneCLI is an open-source credential injection gateway specifically responsible for injecting real credentials on behalf of AI Agents
5 more timeline events
All Facts (13)
Agent frameworks that emphasize security, controllability, and self-hosting are likely to become an important component of enterprise AI infrastructure
65%UnverifiedOneCLI is an open-source credential injection gateway specifically responsible for injecting real credentials on behalf of AI Agents
60%UnverifiedEach AI agent in OneCLI runs in an isolated sandbox environment, with task execution confined to a controlled scope.
50%UnverifiedOneCLI provides a policy-based control system that allows enterprises to granularly manage agents' operational permissions, accessible resources, and approval actions.
50%UnverifiedOneCLI adopts a zero-credential-touch design, where agents never directly access real credentials, reducing the risk of credential leakage at the architectural level.
50%UnverifiedSandbox isolation design can limit the impact scope when an agent encounters a malicious prompt injection attack
50%UnverifiedOneCLI's self-hosted mode allows enterprises to deploy the entire agent framework within their own infrastructure, keeping data within the internal network, and leveraging its open-source nature for code-level transparency and auditability
50%UnverifiedOneCLI targets medium-to-large teams with security and compliance requirements that wish to introduce AI assistants at scale
50%UnverifiedBy ensuring that keys never enter the Agent's possession at any point, OneCLI achieves a complete separation of 'usability' and 'visibility'
50%UnverifiedOneCLI matches based on the target host and path, decrypts and injects the real credentials, and then forwards the request to the external service
50%UnverifiedOneCLI cannot replace the principles of least privilege, key rotation, and auditing; credential lifecycle management remains the responsibility of operations
50%UnverifiedThe core mechanism of OneCLI is that real credentials are stored in OneCLI first, the Agent only receives a fake key, and the gateway injects the real credentials on demand
50%UnverifiedOneCLI 将自身定位为面向团队的开源沙箱化 Agent harness（智能体运行框架）
50%