HybridClaw

HybridClaw is a lightweight AI agent runtime for running AI agents in production. It is described as a self-hosted open-source runtime or a managed cloud service in the EU, and it focuses on the operational side of agentic work: business skills, controls, observability, and deployment.

The runtime includes 35+ verified business skills and 17 built-in tools, each with evals, a security scan, and a scorecard. It supports browser automation, cron workflows, and subagent delegation, and agents can be reached through a terminal UI. The system is also described as handling company knowledge, tool access, permissions, and EU hosting. Its monitoring and telemetry layer includes dashboards and KPIs, with metrics for latency, cost per task, eval scores, and safety incidents. The page also mentions reliability and resilience, cost control, security and audit, scaling, and built-in evals and scorecards.

HybridClaw is built for business workflows. The page names finance, sales, support, operations, BI and analytics, IT and admin, and marketing ops as use cases. It says agents can work over Discord, Teams, iMessage, WhatsApp, E-mail, Web, and Terminal, and that they can operate on websites and tools, work together on tasks, and execute jobs automatically according to a plan or trigger. Examples shown include finance tasks such as collecting invoices and preparing handoff to DATEV, Lexware, or SAP; sales tasks such as researching leads and updating CRM; support tasks such as analyzing tickets; BI tasks such as asking questions of data and creating reports; operations tasks such as running browser workflows; and IT/admin tasks such as checking systems, reviewing logs, and requesting approvals.

HybridAI is named as the management and security layer around the runtime, handling company knowledge, permissions, secrets, audit logs, observability, telemetry, budgets, and deployment. The page also states that passwords and credentials are encrypted, that actions are logged, that sensitive data can be redacted, and that human approvals are available for critical actions. It describes the product as open source and says it avoids vendor lock-in. The page also references 100+ LLM models and business tools including Salesforce, HubSpot, Google Workspace, Microsoft 365, GitHub, Jira, Confluence, and Twilio Voice.