getunbound.ai

What It Is

Unbound positions itself as an Agent Access Security Broker (AASB) for AI coding agents. Rather than a traditional AI Gateway, it acts as a control layer for governing what AI agents can “see, touch, and do” across developers’ local and engineering environments. It covers scenarios involving Cursor, Claude Code, GitHub Copilot, Cline, Windsurf, Roo Code, Gemini CLI, and MCP servers.

Core Capabilities

In terms of protection coverage, Unbound focuses on terminal command execution, MCP connection sprawl, sensitive data exfiltration, credential inheritance, auto-approve drift, and the spread of Shadow AI tools. It can scan AI coding agents, MCP servers, sub-agents, IDE plugins, and risky configurations across an organization, then produce risk posture scores by developer, team, and organization. At the policy layer, it supports semantic parsing of terminal commands, MCP connection controls, and action-level governance, with enforcement modes that can be progressively strengthened from audit to warn, approval, and block. For management and alerting, the product provides full behavior auditing, risk trends, drift detection, approval workflows, and compliance reports for leadership. Enterprise customers can export data to Splunk, Datadog, Elastic, or webhooks.

Deployment, Integrations, and Compliance

Deployment is positioned as requiring no code changes. It can be rolled out through MDM platforms such as Jamf, Intune, JumpCloud, and Kandji, or via a lightweight agent. It also supports CLI scanning, Hooks, and Gateway mode, making it suitable for starting with visibility and inventory before moving into policy enforcement. On compliance, the website explicitly states that Unbound is SOC 2 compliant and can provide evidence for SOC 2, ISO 27001, and regulatory audits; beyond SOC 2, it does not disclose additional certifications of its own. Enterprise features include SAML/OIDC, SCIM, RBAC, environment isolation, CI/CD, and custom integrations.

Pricing and Value

The Pro plan costs $10/user/month, billed annually, with no minimum number of users. It includes 5,000 pooled tool use evals/user/month, a policy engine, 30-day audit logs, and a 48-hour email SLA. Enterprise starts at $18/user/month and targets organizations with 100+ developers. It adds SSO, RBAC, SCIM, SIEM export, 90+ days of logs, a dedicated CSM, Slack Connect, and a 4-hour P1 SLA, while also charging a 10% API Governance Fee on LLM passthrough. For teams already using AI coding tools at scale, the pricing is reasonably aligned with the value of risk governance.

Pros, Cons, and Best Fit

Its main strength is precise positioning: it fills the visibility gap left by traditional CASB, EDR, IAM, SAST, and AI Gateway products when it comes to real-time IDE/CLI/MCP behavior, while supporting governance across heterogeneous tools. The downside is that AASB is still a relatively new category, so actual coverage depth, false-positive rates, and compatibility with complex development environments should be validated through a PoC. Some enterprise concerns—such as data residency, payment methods, and additional compliance details—are not disclosed. Unbound is best suited for CISOs, AppSec teams, platform teams, and organizations with 100+ developers.

Access from China

The website does not state whether direct access, payment, or procurement is available from mainland China, so china_access can only be assessed as unknown. If access or procurement is restricted, organizations can first use existing SIEM, DLP, EDR, IAM, code scanning, and AI Gateway tools as a layered substitute, but these only cover part of the capability set and are unlikely to fully replace Unbound’s AASB use case.