Enterprise open source
Agentic firewalling that keeps autonomy safe.
Clawgress is a purpose‑built network control plane for AI agents. It enforces DNS egress, policy‑driven allowlists, and observability so autonomous systems can run safely in production.
{
"version": 1,
"allow": {
"domains": ["api.openai.com"],
"ports": [53, 80, 443]
},
"labels": {"api.openai.com": "llm-provider"}
}Built for agentic systems, not retrofitted.
Traditional firewalls assume humans at the keyboard. Clawgress assumes autonomous agents. It provides deterministic egress policy, DNS‑level enforcement, and a clean control plane for AI runtime safety.
Policy‑first egress
Allowlist domains, ports, and IPs with a single JSON policy.
DNS enforcement
RPZ‑backed DNS control with explicit deny reason logging.
Observability
Visibility into denies, labels, and applied policy state.
What ships in MVPv1.
Why agents need a firewall.
Agents can call tools, reach APIs, and explore the web. That power requires containment. Clawgress provides the missing layer of control: deterministic, auditable egress built for autonomous systems.
Least‑privilege by default
Ship with a strict baseline policy and expand only when required.
Explainable enforcement
Each deny is labeled so operators can reason about intent and risk.
Ops‑friendly
Standard Linux tooling with predictable network behavior.
Deploy in minutes.
Download the ISO/OVA/QCOW2 image.
Boot and apply a minimal policy.
Route agent egress through Clawgress.
We ship OSS with enterprise expectations.
Documentation
Quickstart, policy schema, and troubleshooting guides.
Community
Public roadmap, issues, and transparent progress updates.
Commercial support
Need SLAs or custom policy work? We can help.
Let’s talk.
General inquiries:
hello@clawgress.ioSecurity & responsible disclosure:
security@clawgress.ioPress:
press@clawgress.io