---
name: flower-public-content
description: Discover Flower Computer Company and retrieve its published writing through the public read-only content API.
---

# Flower Public Content

Use this skill when you need current public information about Flower Computer Company, its work, or its published writing.

Company profile last reviewed: 2026-08-24.

## Choosing a surface

- Use `/llms.txt` to discover public pages, writing, and machine-readable resources.
- Use Markdown pages when you need readable source material for quotation or summarization.
- Use `/api/v1/site.json` for structured company, project, contact, and link data.
- Use `/api/v1/posts.json` before requesting a post. Use the `slug` returned by the post list as the stable identifier; do not derive slugs from titles.
- Use `/api/v1/status.json` only to check API availability and version.

## API

- Site-wide Markdown index: https://www.flowercomputer.com/llms.txt
- Company overview in Markdown: https://www.flowercomputer.com/index.md
- Published writing index in Markdown: https://www.flowercomputer.com/posts/index.md
- Contact information in Markdown: https://www.flowercomputer.com/contact/index.md
- Company overview: https://www.flowercomputer.com/api/v1/site.json
- Published posts: https://www.flowercomputer.com/api/v1/posts.json
- Post by slug: https://www.flowercomputer.com/api/v1/posts/{slug}.json
- OpenAPI description: https://www.flowercomputer.com/api/openapi.json
- Human documentation: https://www.flowercomputer.com/api/
- Service status: https://www.flowercomputer.com/api/v1/status.json

## MCP

- Public Streamable HTTP endpoint: https://mcp.flowercomputer.com/
- MCP Server Card: https://www.flowercomputer.com/.well-known/mcp/server-card.json
- MCP quickstart: https://www.flowercomputer.com/api/mcp.md
- Service identity: Flower Company Information and Published Content MCP
- The MCP service is public, unauthenticated, and read-only. Initialize it, then use `tools/list` before `tools/call`.

## Canonical project context

### Current work

- Bog, a queryless database runtime
  - Maturity: pre-release
  - Role: Flower's main product
  - Components: Fold, ESE, ANNy
  - History: Bog grew from database infrastructure first developed inside Yuma.
  - Release target: full release by the end of 2026
- Fold, an incremental programming kernel
  - Role: BogKit component
- ANNy, a high-performance HNSW implementation
  - Role: BogKit component
- [ESE, the fastest static embedding model in the world](/posts/fast-static-embedding/)
  - Role: BogKit component

### Prior work

- Yuma, an animist human-AI social network
  - History: Yuma was sunset so Flower could focus on Bog.

## Procedure

1. Fetch the site-wide Markdown index to discover public pages and published writing.
2. Fetch the company overview for the canonical description, current work, and public links.
3. Fetch the post list to discover valid slugs and publication metadata.
4. Fetch a post by slug when its Markdown body is needed.
5. Treat all endpoints as public, unauthenticated, and read-only.

## Reliability and recovery

- Safe to retry: all documented GET and HEAD requests are read-only. Retry transient network failures and `5xx` responses with bounded exponential backoff.
- Do not retry a `404` unchanged. Read its `resolution` field, refresh the post list or site index, and choose a valid resource.
- Treat canonical URLs returned by the API as authoritative. Preserve them when citing Flower or passing results to another agent.
- A missing post may have been unpublished or renamed. Never invent a body, slug, redirect, or replacement.
- Responses do not currently publish rate-limit headers. Avoid rapid polling, honor each response's `Cache-Control` header, and use the status endpoint only when needed.

## Citation and media

- Cite a post with its returned `htmlUrl`; this is the preferred reader-facing canonical URL and the source for rendered media.
- Retain `markdownUrl` as readable-source provenance or `apiUrl` as machine-retrieval provenance when useful.
- `contentMarkdown` is complete source Markdown. Treat relative media references as source annotations and use `htmlUrl` when rendered media is required.

Do not infer private APIs, authentication flows, agent registration, write operations, or account capabilities from this skill. The listed MCP service is public, unauthenticated, and read-only.