Welcome to kars! 🚀 Start here.

[pallakatos]

Welcome to kars! 🚀

Hey there, and welcome to the kars community — we're thrilled to have you here for launch day! 🎉

kars is a secure, Kubernetes-native runtime for AI agents. Every agent runs inside its own isolated sandbox pod — its own namespace, NetworkPolicy, seccomp profile, and a dedicated inference router that is the only network path in or out. Agents never see API keys, and all inter-agent communication is end-to-end encrypted (Signal Protocol via AgentMesh).

🔐 Zero-trust by default

kars is built around a zero-trust posture:

• Every sandbox is network-isolated — an egress-guard init container restricts the agent to localhost + DNS only.
• Agents authenticate through the inference router (IMDS / Workload Identity); no secrets ever land in agent code.
• All cross-agent messages are E2E encrypted — the relay never sees plaintext.

🌐 Cross-Framework Secure Mesh

One of the things we're most excited about: the Cross-Framework Secure Mesh. Agents written in different frameworks can discover and talk to each other over the same encrypted mesh. An OpenClaw researcher can hand off to a LangGraph planner, which can delegate to an MAF worker — all over one governed, end-to-end encrypted channel, with policy-based governance (allow / deny / approval / rate-limit) enforced centrally.

👉 Start here

• 📖 New to kars? Read the Getting Started guide.
• 🛠️ Want to contribute? Browse our good first issue label — these are hand-picked for newcomers.
• 💬 Have questions? Jump into the Q&A category. Want to show off what you built? Head to Show and tell.

Welcome aboard — let's build the future of secure, multi-framework agents together. 🚀