Skip to content
Memstead Docs

Memstead documentation

Memstead gives AI agents a typed, validated model of a project — plain markdown in git, validated on every write against a schema you control. New here? Start with the tutorial, then see Skills for what you actually run day to day — before diving into the concept and reference layers.

Guides

Section titled “Guides”
  • Getting started — install, memstead quickstart, first entities, agent connect.
  • Skills — the eight-skill plugin roster as a job → command map: what to run for setup, capture, building, syncing, verifying.
  • Author a schema — scaffold a custom schema and pin a mem to it.
  • Publish a mem — share a mem through the memstead.io registry.
  • Agent recipes — worked MCP tool-call sequences with real payloads.
  • Back up a mem-repo — push mems to any git remote and recover them on another machine.

Concepts

Section titled “Concepts”
  • Glossary — normative definitions: mem, schema, workspace, mount, entity, and the rest. Built from the repository’s GLOSSARY.md on every deploy.
  • The fidelity contract — what a verify run measures (coverage, accuracy, freshness), how it reports honestly, and what it deliberately does not measure yet.
  • Vision — what Memstead is for and the design rationale.

Reference

Section titled “Reference”

Generated deterministically from the live source on every push: every MCP tool, every CLI subcommand, every UniFFI method, every typed error code that the engine, the CLI, the MCP servers, and the Registry HTTP envelope produce. The Surface Parity Matrix lines them up so cross-surface gaps are visible by inspection.

Generated fromdevonunbuilt