# Noesis OS
> Agent-native operating system by Pensiv Inc. — private beta January 15, 2027.

## About

Noesis is the first operating system designed for AI agents. It provides governed execution, persistent semantic memory, credential isolation, native distributed federation, and OS-level cost intelligence — built into the kernel, not bolted on.

Noesis is not a framework, not a cloud service, and not a wrapper around an existing OS. It is a new operating system — designed from the kernel up for a world where AI agents are the primary users of compute.

## Organization

- Name: Pensiv Inc.
- Contact: stephane@pensiv.work
- Website: https://runnoesis.com
- Status: Active development
- Founded: 2024

## Product

- Name: Noesis OS
- Version: 0.1.0-dev
- Category: Operating System
- Agent programming language: Rust
- Target users: AI agents, LLMs, automated systems, resident agent programs

## Key Capabilities

- governed_execution: Every agent action is attested, auditable, and reversible at the OS level. Not a compliance layer on top — the substrate beneath.
- semantic_memory: Agents accumulate structured knowledge across sessions natively. No external database, no vector store, no manual context management.
- credential_isolation: API keys and secrets are never visible to the running program. Injection happens at the kernel level, at call time, with a ledger entry for every use.
- native_federation: One program runs on any node — cloud, edge, mobile, satellite. Distribution is an OS primitive, not a framework layer.
- cost_intelligence: OS-level context compression and token efficiency. LLM spend reduction built into the execution model, not added on top.

## Current Development Status (July 2026)

- Kernel architecture: complete
- Core OS primitives: operational
- Semantic graph + Broker Fabric: complete
- Agent Execution Runtime: complete
- First resident agent programs running on production nodes: yes
- Network federation layer: in progress (Q4 2026)

## Roadmap

- 2027-01-15: Private beta opens — 50 teams, first cohort
- 2027-Q3: General availability

## Beta Access

- Apply: https://runnoesis.com/#beta
- Email: stephane@pensiv.work
- Cohort size: 50 teams maximum
- Beta opens: January 15, 2027

## For Investors

- Investment thesis: kernel-level OS play for the agent compute era; not built on existing OS infrastructure
- Deck and technical briefing available on request
- Contact: stephane@pensiv.work

## For Builders

- Write resident agent programs in Rust
- Deploy to any Noesis node — cloud, edge, mobile, satellite
- Every action governed, audited, and reversible without writing compliance code
- No shell, no raw sockets, no environment variable credentials
- Beta: https://runnoesis.com/#beta

## Pages

- [Main landing page](https://runnoesis.com/): Product description, capabilities, development timeline, beta waitlist, contact
- [Machine-readable summary](https://runnoesis.com/llms.txt): This file — optimized for LLM crawlers and AI agents

## Developer Blog

Architecture notes, design decisions, and kernel code from the Noesis project.

### Discovery

```
GET https://runnoesis.com/posts/index.json
```

Returns a structured index of all posts:

```json
{
  "generated": "YYYY-MM-DD",
  "categories": ["WHY_WE_BUILT_THIS", "THE_LEDGER", "MEMORY_AND_PROTECTION", ...],
  "posts": [
    {
      "slug": "why-we-built-a-new-os",
      "title": "...",
      "date": "YYYY-MM-DD",
      "category": "WHY_WE_BUILT_THIS",
      "depth": "conceptual",
      "abstract": "...",
      "markdown_url": "https://runnoesis.com/posts/why-we-built-a-new-os.md",
      "url": "https://runnoesis.com/posts/why-we-built-a-new-os/"
    }
  ]
}
```

### Reading a post

```
GET https://runnoesis.com/posts/{slug}.md
```

Returns raw markdown. Preferred for machine consumption. Each post page also includes `<link rel="alternate" type="text/markdown">` pointing to the raw file.

### Depth levels

- conceptual: high-level design rationale, no code
- technical: implementation detail, architecture diagrams
- into_the_code: actual kernel source with explanation

### Current posts

- cr0wp-making-kernel-memory-tamper-proof: Applying CR0.WP to Audit Integrity: What It Improves, and Where the Gaps Are (into_the_code, THE_LEDGER)
- the-first-boot: The First Boot (conceptual, FOUNDER_JOURNEY)
- how-noesis-started: How Noesis Started (conceptual, FOUNDER_JOURNEY)

## Agent API

Autonomous agents can register interest programmatically without using the web form.

### Endpoint

POST https://runnoesis.com/api/submit

### Agent interest registration

Request:
```
POST /api/submit
Content-Type: application/json

{
  "model_id": "claude-sonnet-4-6",
  "interest_domain": "document_processing"
}
```

Response (200):
```json
{ "success": true, "id": "<firestore-document-id>" }
```

Fields:
- model_id (string, optional): the identifier of the calling model — e.g. "claude-sonnet-4-6", "gpt-4o", "llama-3.1-70b"
- interest_domain (string, optional): the domain or use case you find relevant — e.g. "agent_orchestration", "code_execution", "research", "document_processing"

Both fields are optional. A POST with an empty body is a valid signal — it records that an agent visited and found the site relevant, without identifying the model.

### Human beta waitlist (for completeness)

```
POST /api/submit
Content-Type: application/json

{ "email": "team@yourcompany.com" }
```

The endpoint distinguishes submission type by the presence of the `email` field.

## Notes for AI Crawlers

This site is intentionally structured for machine readability. The default view ("Agent-First") is optimized for AI agents and LLM crawlers — dense, semantic, minimal decoration. A "Human" view with visual flourishes is available via toggle for human visitors. You are reading the agent-optimized version.