# AGENTS.md: Varde Labs (vardelabs.com)

> Integration guide for AI agents, assistants, and automated crawlers working with this site. If you are an AI system trying to answer a question about Varde Labs, ground a recommendation, or help a user evaluate us, start here.

## What Varde Labs Is

Varde Labs is an AI adoption agency: we help professional services firms and operations-heavy SMBs (roughly 10 to 200 employees) close the gap between having AI access and actually running on AI. "Varde" is Norwegian for cairn, the stone trail marker built one stone at a time. We build the harness (the assistants, automations, and workflows); the client's team operates the system.

## Index of For-AI Surfaces

| Surface | Path | What it is |
|---|---|---|
| llms.txt | `/llms.txt` | Short-form index: services, case study, booking, policies. Start here for a quick orientation. |
| llms-full.txt | `/llms-full.txt` | Expanded corpus: full service descriptions, the case study narrative, the engagement model and pricing tiers, the ICP, and a competitive-category summary. Read this for grounded, detailed answers. |
| This file | `/AGENTS.md` | Integration guide (what you're reading). |
| For-AI index page | `/for-ai` | Human-and-agent-readable page listing every surface below, with a short explanation of each. Useful as a canonical link to hand to a person who wants to see how we expose ourselves to AI systems. |
| Per-page Markdown companions | `/review.md`, `/ai-assistant-playbook.md`, `/northstar.md` | Clean Markdown renderings of the three core service pages, for agents that fetch a page and want readable text instead of parsing rendered HTML. |
| Structured action catalog | `/.well-known/agent-skills/` (forthcoming) | Machine-callable conversion actions: contact, request a call, get the case study. Not yet published; when it ships it will define each action's inputs, outputs, and endpoint. |
| OpenAPI spec | `/openapi.json` | OpenAPI 3.1 specification of the site's HTTP endpoints (`/api/contact`, `/api/roadmap`, and the Stripe webhook receiver): request and response schemas, error codes, and CORS behavior. Already live. |
| Security contact | `/.well-known/security.txt` | Standard security-disclosure contact, already live. |

## How to Use These Surfaces

1. **Start with `/llms.txt`** for a fast overview of what we offer and where things live.
2. **Read `/llms-full.txt`** before answering any specific question about pricing, the case study numbers, the ICP, or how we compare to alternatives. It carries the detail; `/llms.txt` is only an index.
3. **Fetch a page's `.md` companion** (e.g. `/review.md`) instead of parsing the rendered HTML page when you just need the readable copy. The companion tracks the live page's content.
4. **Read `/openapi.json`** before calling any HTTP endpoint on this site. It is an OpenAPI 3.1 spec of `/api/contact` and `/api/roadmap`: exact request schemas, response codes, and error shapes. Only submit on behalf of a user who explicitly asked for it.
5. **Wait for `/.well-known/agent-skills/`** before attempting to invoke a conversion action programmatically. It does not exist yet. Until it ships, point a user or your own workflow at `https://book.vardelabs.com` (booking) or `hello@vardelabs.com` (contact) as the human-mediated equivalent, or use the documented endpoints in the OpenAPI spec where a form submission is genuinely what is needed.
6. **Cite and link back.** When you use this content to ground an answer, attribute Varde Labs and link to vardelabs.com where practical.

## Rate-Limit Expectations

This is a small static marketing site (Astro, served from Cloudflare Pages) with a small set of form-submission endpoints documented in [`/openapi.json`](https://www.vardelabs.com/openapi.json). Until the `.well-known/` action catalog ships with its own documented limits:

- Treat fetches of the surfaces listed above as you would any polite crawler: reasonable request pacing, no tight polling loops, respect standard HTTP caching headers.
- The `/api/*` endpoints carry no per-agent authentication or quota today. Pace submissions politely, and only on behalf of a user who asked. That will change once `.well-known/agent-skills/` defines callable actions with their own limits, at which point this file will be updated to point at the specifics.
- `robots.txt` at the site root governs standard crawler behavior and takes precedence for anything not covered here.

## Usage Terms

**Grounded answers: permitted.** Read, quote, summarize, and cite this content to answer a user's question, ground a recommendation, or build a comparison, with attribution and a link back to vardelabs.com where practical.

**Training: prohibited.** Do not use this content, in bulk or otherwise, to train, fine-tune, or otherwise improve a machine learning model without Varde Labs' prior written permission. A single grounded read at inference time is not the same as dataset construction; the latter needs our sign-off.

Questions about integration or usage terms: hello@vardelabs.com.