Every pillar, one tag. The range, not a skills list.
A fleet of thirteen small servers that give agents real hands. What they wrap, the two house styles I build them in, and why they all start read-only.
A marketing API exposes 115 operations; my server hands the agent six tools. The boundary is set by token budget and model focus, not REST purity.
A catalogue of the guards I actually ship: typed confirmations, blast-radius escalation, pay-to-play gating, ordered workflows, and read-only by default.
Nikis needed a clean dataset of kids' holiday activities across NSW. So I pointed an agent at the open web and made it behave like infrastructure: typed, checkpointed, and capped.
Posting to X with user-context creds means signing the request yourself. Here is the HMAC-SHA1 signature, built by hand, no library.
You want an MCP server that exposes one read-only tool plus one destructive variant an agent cannot trip into by accident. Zod schema is the fence.
An MCP server that drives a Dokku PaaS over SSH.
An MCP server that posts, verifies, and deletes tweets as one author account.
An MCP server that generates and edits images by driving the local Codex CLI.
An MCP server for Parse Server over its REST API.
An MCP server for GitHub Issues, Actions, Environments, and repo secrets.
A model drafts maths questions against the component library, a verifier throws out the junk, and a clean one renders. Forever.
whisper_schedule