Clawvisor provides an AI‑agent security gateway that sits between an autonomous agent and the large language model or external tools it invokes. By routing every request through a configurable policy layer, it verifies identity, scopes tool access, and logs activity before any call reaches the model, enabling organizations to enforce security without modifying agent code. The open‑core solution can be self‑hosted or managed and integrates with any agent framework and popular services such as GitHub, Gmail, and AWS.
Funding
Funding not disclosed
Founders
Product
Problem
AI agents need access to external tools and credentials to be useful, but granting them unrestricted secrets creates a high risk of accidental or malicious misuse, data leaks, and compliance violations.
Solution
Clawvisor provides a drop‑in security gateway that sits between an AI agent and the large language model or external APIs it invokes. By routing every model and tool request through the gateway, Clawvisor verifies the task’s purpose, scopes short‑lived credentials, and logs the action before any call is executed. The gateway injects real secrets only at call time, keeping them hidden from the agent and the LLM. Policies can be defined with RBAC/ABAC rules, and high‑risk actions can be held for human approval, delivering defense‑in‑depth without requiring code changes or SDK integration. The core gateway is open source and can be self‑hosted, while managed cloud options add shared policy, risk scoring, and support for teams.
Target Audience
Primary customers are engineering teams that deploy production AI agents—such as developers building autonomous assistants, workflow automation bots, or AI‑driven integrations—who need secure, auditable tool access and credential protection.
Features
- Transparent drop‑in deployment: point the agent’s base URL to the gateway, no SDK or code modifications required
- Purpose‑based authorization that evaluates each request against declared task intent and policy rules
- Scoped, short‑lived credential handles injected at call time; real secrets remain in a vault and are never exposed to the agent or model
- Full audit trail of every tool and model call with per‑task containment and automatic revocation on task completion
- Risk scoring and optional human review for high‑impact actions such as data deletion or financial transfers
- Open‑core implementation: self‑hostable gateway with community support or managed cloud service with additional governance features
- Role‑based access control, shared policies, and extended log retention for team and enterprise plans