KruxOS¶
The operating system built for AI agents.
KruxOS replaces the traditional Linux command line with typed, documented, governable APIs. Every OS operation — reading files, running processes, sending emails, managing secrets — is a structured capability that agents discover, invoke, and get structured results from.
No shell parsing. No permission guessing. No silent failures.
Who is KruxOS for?¶
-
:material-account:{ .lg .middle } Users
Get KruxOS running, connect your AI agent, and start using it in under 15 minutes.
-
:material-shield-lock:{ .lg .middle } Enterprise
Evaluate KruxOS for production: security model, compliance, architecture, and benchmarks.
-
:material-code-braces:{ .lg .middle } Developers
Build capability packs, use the Python SDK, and contribute to KruxOS.
Why KruxOS?¶
The problem¶
AI agents running on traditional Linux spend 40-60% of their tokens parsing shell output, handling text errors, and guessing at permissions. Every command is a string. Every response is unstructured text. Every error is a surprise.
The solution¶
KruxOS exposes every OS operation as a typed, documented API:
| Traditional Linux | KruxOS |
|---|---|
ls -la /workspace (parse text output) |
filesystem.list(path="/workspace") (structured JSON) |
echo $? (check exit code) |
Structured error with type, description, recovery actions |
chmod 755 file (hope it works) |
Policy engine evaluates permission in real time |
grep -r "password" .env (secrets in plaintext) |
Vault injects secrets at runtime, agent never sees raw values |
| No audit trail | Every operation hash-chained to append-only audit log |
Key features¶
- 89 typed capabilities across 13 categories — filesystem, process, network, git, scheduler, system, agent, state, comms, secrets, email, Slack, alerts. MCP-native (
tools/list) with JSON-RPC fallback (capabilities.list). - Deterministic policy engine — four-tier model (
autonomous/notify/approval_required/blocked), hot-reloadable YAML at/data/kruxos/policies/, per-agent overrides, no LLM in the hot path. - One approval surface — Claude Code, OpenAI Codex, and any cli-config'd CLI route every gated tool call through the KruxOS queue. No in-CLI prompts; no governance bypass via native shell tools.
- Per-agent sandbox — Linux user/network namespaces, cgroup v2 limits (512 MiB memory / 50% CPU / 100 PIDs / 50 MiB/s read / 25 MiB/s write defaults), seccomp BPF allowlist, nftables defense-in-depth.
- Service Proxy — read-replica + write-buffer + batch-protection chain for Gmail and Slack (adapters ship in v0.0.1; operator-facing OAuth UX lands in v0.0.2).
- Multi-agent runtime — five-field cron schedules, one-shot delays, manual trigger via
kruxos agent run, topic-based inter-agent comms broker, three state scopes (session / persistent / shared). - Secrets vault — AES-256-GCM SQLite. Use-not-read contract: capability handlers reference vault entries by id, never the raw secret.
- Hash-chained audit — length-prefixed CBOR, Principal-aware actor field, daily rotation, 90-day retention default.
- Web dashboard — real-time supervision, approval queue, audit viewer, multi-model chat with persisted sessions, visual + YAML policy editor, first-boot wizard, code sessions (VM image only in v0.0.1).
- Model-agnostic — Anthropic, OpenAI (+ Codex + DeepSeek / Grok / Mistral / Groq / GLM via
base_url), OpenRouter (200+ models), Google Gemini, Local (Ollama / vLLM / LM Studio / llama.cpp).
Quick start¶
docker run -d --name kruxos --privileged \
-e KRUXOS_VAULT_PASSPHRASE='choose-a-strong-passphrase' \
-p 7800:7800 -p 7700:7700 -p 7701:7701 \
-v kruxos-data:/data/kruxos \
altvale/kruxos:latest
Open https://localhost:7800 and complete the first-boot wizard (vault passphrase, AdminAgent, license, User token, CLI install).
Download the .qcow2 / .vmdk / Vagrant .box / raw .img.gz for x86_64 or aarch64, boot, and complete the dashboard wizard. All artefacts cosign-signed with offline-verifiable .cosign.bundle.
Architecture at a glance¶
graph TB
Agent[AI Agent] -->|MCP / JSON-RPC| Gateway[Agent Gateway :7700]
Gateway --> Registry[Capability Registry]
Gateway --> Policy[Policy Engine]
Registry --> Capabilities[89 Capabilities]
Registry --> Proxy[Service Proxy]
Proxy --> Gmail[Gmail]
Proxy --> Slack[Slack]
Policy --> YAML[YAML Rules]
Gateway --> Vault[Secrets Vault]
Gateway --> Audit[Audit Logs]
Gateway --> State[State System]
Gateway --> Comms[Agent Comms]
Dashboard[Web Dashboard :7800] -->|Supervision WS :7701| Gateway
CLI[kruxos CLI] --> UserAPI[User API :7703]
UserAPI --> Gateway
Community¶
KruxOS is open source under the Apache 2.0 license. The community edition includes everything you need to run AI agents safely — the web dashboard, all 89 capabilities, the policy engine, and the in-appliance Python SDK.