About Octopus
An AI Agent Operating System engineered from line one for determinism, safety, and auditability
Our Mission
Build the world's first formally correct autonomous AI Agent operating system. Not another GPT wrapper, but a complete OS rigorously verified from kernel to product surface.
Our Vision
Empower every organization with trustworthy, auditable, and predictable autonomous AI Agents—covering everything from IT operations to business workflows, security to knowledge management.
What Makes Us Different
Deterministic Kernel
552 Frozen Contracts552 frozen contracts, 52 compile-time gates. Same input always produces same output. Not a probabilistic black box, but a verifiable engineering system.
Zero I/O Architecture
Zero Side EffectsAbsolute zero I/O in the kernel layer. All external interactions delegated via Port protocols. 1,612 kernel tests run without any external dependencies.
Evidence-Driven
100+ Event TypesEvery decision has a complete evidence chain. 100+ LiveEvent types record operation traces. replay.jsonl supports full historical replay.
Full-Stack Coverage
5 Product FormsFrom CLI to Desktop, Web to Mobile—5 product forms sharing a unified backend. 46+ role definitions covering 15 scenario families.
By the Numbers
Technology Foundation
Kernel / Server / CLI / TUI / Base — five layers
28 route modules, 546 endpoints, SSE streaming
Web Console + Desktop (Tauri) + Mobile (Expo)
Local state + vector semantic retrieval
Agent task scheduling and workflow orchestration
Containerized capability execution, security sandbox
Layered Openness
Octopus adopts a Community Core + Enterprise Governance model: foundation layer fully open under AGPL, governance layer source-available, enterprise core commercially licensed. Openness doesn't mean chaos; transparency doesn't mean vulnerability.
Let's Talk
Whether it's technical discussion, investment partnership, or enterprise customization, we look forward to connecting with you.