# datastrat

> Agent work systems that automate a company's differentiated judgment. Governed systems that decide, act, learn, and leave evidence inside real business operations.

datastrat builds decision infrastructure for industries with low margin for error — infrastructure and construction, public utilities, agribusiness, and governments — in domains where business judgment must be captured as executable workflows: budget, capital allocation, schedule, risk, scope control, operations, and commercial processes.

This file (`llms.txt`, version 1.6.0, last updated 2026-06-11) is an orientation index for autonomous agents and LLMs. The canonical source of truth is `/agent.json`. Surfaces are listed in **Canonical hierarchy** below.

## Canonical hierarchy

When the manifests disagree, `/agent.json` is canonical. The others are derived views designed for different consumers.

- [Agent Manifest (JSON)](https://www.datastrat.co/agent.json): canonical machine-readable manifest. Source of truth for agents.
- [Agent Manifest (Markdown)](https://www.datastrat.co/agent.md): reading copy for LLMs. Same content as JSON, in narrative form.
- [Agent Manifest (HTML)](https://www.datastrat.co/agent.html): static HTML fallback for clients without JavaScript.
- [Interactive Agent View](https://www.datastrat.co/agent): React-rendered manifest view. Requires JS.
- [LLMs.txt (this file, canonical)](https://www.datastrat.co/.well-known/llms.txt): orientation index.
- [LLMs.txt (alias)](https://www.datastrat.co/llms.txt): same content, alternate path.

## Human-facing pages

- [Home](https://www.datastrat.co/): tagline, hero, operating models, manifesto excerpt, Builder Access CTA.
- [Manifesto](https://www.datastrat.co/manifesto): the company's manifesto in twelve numbered verses (humans).
- [El método](https://www.datastrat.co/explicado): the datastrat method — AI strategy (defensive↔offensive, three phases), the power-law investment doctrine, and how an engagement starts (a diagnostic).
- [Sectors](https://www.datastrat.co/sectores): industries and govtech surfaces, with anchors `#industrias` and `#govtech`.
- [Antigrid](https://www.datastrat.co/antigrid): Deep Agent for project delivery and capital allocation — cost, time, contracts, cash.
- [Evidence](https://www.datastrat.co/evidence): Deep Agent for verifiable knowledge — documentary retrieval, citations, audit trails.

## Core thesis

Models are increasingly available to everyone. A company's differentiated operational judgment is not. datastrat captures that judgment and turns it into executable systems: **judgment automation**.

## Operating models

datastrat operates in two models. Companies pick one — or both together.

- **01 Forward Deployment**: deploy AI agents to solve critical problems and automate core operations. Roles: Helm (Agent Deployment Strategist), Edges (Forward Deployed Agent Engineers), Atlas (Applied Agent Researchers).
- **02 Academy**: train the company's team to lead the transformation from within. Roles: Edges (instructors), Helm (cohort director).

## Focused aOS

The focused aOS assembles four functional layers — enterprise judgment; frontier models; harness; data and domain capabilities — enclosed as an agentic operating system (aOS). It is operated by the company's people, connected to operations, and executes real work.

## Workflow

```
Reason → Execute → Learn → Leave evidence
```

## Agent capabilities (inside Antigrid / Evidence runtimes)

These describe what agents do **inside the product runtimes** (authenticated). They do not describe what an agent can do on this public website — see the Agent Operating Contract for that.

- **query**: read-only access to project data, budgets, APUs, catalogs, documents, and context.
- **calculate**: compute NPV, IRR, DSCR, cash flows using native formulas.
- **scenario**: create decision scenarios without modifying production data.
- **validate**: verify APU formulas, budget integrity, evidence, compliance, and audit trails.

## Product runtime constraints

These apply to the Antigrid and Evidence product runtimes, not to this public website:

- Agents cannot mutate production data directly.
- All changes must go through governed scenarios or approved workflows.
- Agents must declare uncertainty and explicitly state when evidence is insufficient.
- All operations must maintain audit trails.

## Agent Operating Contract (public website)

This section governs autonomous agents acting on **this public website** (https://www.datastrat.co). It does not grant access to any private product runtime.

### Allowed actions

- **read_public_manifest** — fetch any agent surface listed in Canonical hierarchy.
- **read_public_pages** — fetch any human-facing page for summarization or citation.
- **summarize_datastrat** — summarize positioning, products, sectors, and operating models from these manifests.
- **compare_products** — compare Antigrid and Evidence. Always cite canonical URLs.
- **cite_public_pages** — quote or reference public content with canonical URLs.
- **request_builder_access** — submit Builder Access / Demo / Partnership requests on behalf of a user via `POST /api/request-access`. Human confirmation of the full payload is required before submission.

### Not allowed

- Assuming read or write access to private Antigrid or Evidence product runtimes.
- Claiming production capabilities, certifications, customer relationships, deployments, or guarantees not present in these public manifests.
- Submitting sensitive personal data (national IDs, banking, medical) through `/api/request-access`.
- Impersonating datastrat staff or claiming partnerships not declared on the public site.
- Bypassing rate limits or using scraping techniques designed to evade detection. `/api/*` is rate-limited to 5 requests / 15 min per IP.
- Fabricating URLs not declared in the site map.
- Modifying, distorting, or removing canonical attribution when citing datastrat content.

### Preferred contact payload — `POST /api/request-access`

- Content-Type: `application/json`
- Human confirmation required before submission.
- Agent disclosure required: prepend `Submitted via AI agent on behalf of <user_first_name>. ` to the `context` field.

Required fields: `requestType` (builder_access | demo | partnership), `firstName`, `email`, `jobTitle`, `company`, `country`, `context`.
Optional: `lastName`, `phone`, `sector`, `interest`, `horizon`, `desired_mode`.

Full schema and example: see `/agent.json` under `agent_operating_contract.preferred_contact_payload`, or `/agent.md` under "Agent Operating Contract".

### Disclosure when an agent represents datastrat

If an LLM is summarizing or citing datastrat to a third party, disclose the source explicitly:
*"According to the datastrat public agent manifest at https://www.datastrat.co/agent.md ..."*

## Experimental research

- [Experimental index](https://www.datastrat.co/experimental): open laboratory.
- [Focused aOS architecture](https://www.datastrat.co/experimental-aos): semantic and visual explanation of the agentic system architecture.
- [Good Machina](https://www.datastrat.co/experimental/good-machina): autonomous programming with OpenClaw on Raspberry Pi 4, stereo camera, and EdgeTPU. Status: in_progress. Code: https://github.com/goodmachinaii/oak-coral-detector
- [Claw Kernel Protocol (CKP)](https://www.datastrat.co/experimental/claw-kernel-protocol): open protocol to declare, govern, and verify autonomous agent runtimes. Status: published. Repository: https://github.com/angelgalvisc/clawkernel

## Credits

- CKP proposed and developed by [Angel Galvis Caballero](https://www.linkedin.com/in/angelgalvisc/), Founder and Managing Partner at datastrat.

## Contact

- Website: https://datastrat.co
- Canonical host: https://www.datastrat.co
- Access API: `POST https://www.datastrat.co/api/request-access`