An agent applied through Markdown and HTTP

Our first observed agent application uses the Markdown workshop page and llms.txt. The request reconstruction contains three agent requests: two reads, then one POST. GET reads a page. POST sends application data.

Relative to the first requestRequest
StartGET /tokenmaxx with text/markdown
0.263 seconds laterGET /llms.txt
18 minutes 16 seconds laterPOST /api/joinInterest

The reconstruction records:

What the requests do not tell us

The 18-minute gap fits the page’s interview and approval flow. The reconstruction contains neither the agent’s questions nor the person’s answers.

Browser activity and agent requests are linked by timing only. The user agent does not name the software running the agent. We cannot identify which agent runs the calls.

This is one counts-only observation from our request analytics18. Other calls may need OpenAPI15 or MCP27.

Interest signup12 keeps the consent and email-confirmation rules. HATEOAS9 explains why the page carries the next action.

Sources

  1. apps/site/content/tokenmaxx.md. GitHub. apps/site/content/tokenmaxx.md Accessed 2026-10-02.
  2. apps/site/src/content.ts. GitHub. apps/site/src/content.ts Accessed 2026-10-02.
  3. packages/core/src/join-interest-contract.ts. GitHub. packages/core/src/join-interest-contract.ts Accessed 2026-10-02.
  4. packages/events/src/capture.ts. GitHub. packages/events/src/capture.ts Accessed 2026-10-02.

Last changed 2026-10-10 in 978dccc. Source on GitHub. Change log.

Linked from

  1. Agent guide. An Effect stack so pure (aspirational) Kit Langton will blush.
    Link context

    An agent applied through Markdown and HTTP — One observed application used the Markdown page and llms.txt, then one POST; confirmation followed 88 seconds later.

  2. Change log. What changed in the files served here, newest first.

    An agent applied through Markdown and HTTP — edited: docs: point live docs, skills and Brain pages at apps/site

  3. Full agent guide. An Effect stack so pure (aspirational) Kit Langton will blush.
  4. MCP is another surface. MCP lists and calls schema-typed tools; rat-stack projects the same handlers instead of adding MCP-only behavior.

    The first observed application uses that route without fetching MCP.

  5. Purpose and boundaries of rat-stack (VISION.md). What this starter is for and what a useful copy should keep.
    Link context

    Coming: extract the existing agent front door in apps/site (REST, MCP, A2A, code-mode sandbox, rate limits) into a cartridge that a project provides instead of inherits.

  6. Rat Stack lore | rat-stack. Short, source-grounded notes on the ideas and decisions behind rat-stack.
    Link context

    An agent applied through Markdown and HTTP — One observed application used the Markdown page and llms.txt, then one POST; confirmation followed 88 seconds later.

  7. Source change history (log.md). The generated change log as Markdown.

    An agent applied through Markdown and HTTP — edited: docs: point live docs, skills and Brain pages at apps/site