Open-source team agent harness with sandboxed per-employee assistants, network-layer credential injection, policy enforcement and human approvals outside the model.
OneCLI is an open-core team agent harness built around sandboxed per-employee assistants, with credential injection, policy enforcement, human approvals and identity/audit controls kept outside the model itself.12 The reviewed product surface is broader than a simple secrets gateway: the vendor positions it as an operating layer for team agents that can work with services such as GitHub, Gmail, Notion, Dropbox, CRM systems and Slack while authorization and secret handling remain external to the model.1
The central design choice is useful for agentic workflows: agents can request an action without receiving the underlying credential, while deterministic controls can require approval before sensitive operations proceed.13 Y Combinator corroborates the company and its current team-agent/security positioning, but does not independently establish performance or security outcomes.4
The hosted product offers a seven-day trial with $5 in AI credits and no payment card required.1 The repository uses an open-core model: the core is Apache-2.0 and can be self-hosted for production, while enumerated enterprise components are separately licensed and require an Enterprise subscription for production use.25
This gives technical teams a meaningful low-cost evaluation and self-hosting path, but public evidence does not establish complete enterprise list pricing. Buyers that need the enterprise-only components should therefore confirm subscription scope and total cost directly with the vendor.
OneCLI publishes source code, an explicit enterprise-license boundary and a security-reporting process, which provides more inspectable trust material than a closed product with only marketing claims.256 Its current site describes SOC 2 Type II as in progress and GDPR DPA readiness; those statements should not be read as completed independent certification.1
The vendor also describes encrypted vault storage and a network-layer approach intended to keep raw credentials out of model context.3 That architecture is directionally strong, but the reviewed evidence does not include an independent audit demonstrating that the implementation consistently achieves the vendor's strongest security claims in production. The canonical source panel yielded no exact-product independent numeric rating or benchmark evidence.
OneCLI fits engineering, IT and security-conscious teams that want to deploy employee or departmental agents while centralizing credentials, approvals and policy outside the model. The self-hostable core is particularly relevant to teams that want infrastructure control, while the hosted trial lowers evaluation friction.12
It is less compelling for buyers that require mature independent reliability data, completed third-party compliance attestations before evaluation, or fully transparent enterprise pricing. Non-technical teams may also depend on administrators to configure the security and integration layer correctly.
the separation of secrets from model context, deterministic approval controls, per-user agent isolation, an inspectable open-source core and a clear licensing boundary between core and enterprise components.125 The product also addresses a real weakness of many agent deployments: giving useful automation access to external systems without simply placing durable credentials in prompts or agent environments.
early-stage maturity, limited independent evidence, incomplete public enterprise pricing and reliance on vendor assertions for several security/compliance claims. SOC 2 Type II remains in progress on the reviewed surface, so organizations with strict procurement requirements should verify current certification and deployment controls before production rollout.1
Products matched by shared tasks, audience, workflow, product type and capabilities.