Context Sync

Render approved project memory into the files that coding agents and editors already inspect.

What it writes

Context Sync writes managed blocks to:

  • AGENTS.md
  • CLAUDE.md
  • .github/copilot-instructions.md
  • MEMORY.md
  • .cursor/rules/memoryguard.mdc

These are the same files your coding tools already read. MemoryGuard does not introduce a new file format; it keeps the familiar ones aligned with approved project memory.

Basic flow

shell
memoryguard remember "Package manager: pnpm."
memoryguard remember "FastAPI was superseded by Flask."
memoryguard sync
memoryguard status

memoryguard sync generates a context diff, writes the pending plan, and applies it in one step. memoryguard context generate shows the diff and waits formemoryguard context approve if you want a two-step flow.

Watch for drift

memoryguard context watch polls the project and either applies the diff (--yes) or prints a pending diff for you to approve.

shell
# print a pending diff whenever memories change
memoryguard context watch

# apply immediately when changes are detected
memoryguard context watch --yes --once

Supersession

Projects change. MemoryGuard is designed to preserve that change instead of hiding it. When a newer approved memory replaces an older one, generated context can show the current fact while keeping useful history such as “FastAPI was previously used; superseded by Flask.”

Mark a superseded relationship explicitly with memoryguard resolve:

shell
memoryguard resolve <old_memory_id> --superseded-by <new_memory_id>

Managed blocks

MemoryGuard writes managed context blocks bounded by <!-- memoryguard:context:start --> and <!-- memoryguard:context:end --> markers. Rerun memoryguard sync whenever the truth of your project changes, and the managed section refreshes in place.

Review before publishing

Treat generated context as code-adjacent output. Review AGENTS.md, CLAUDE.md, .github/copilot-instructions.md, MEMORY.md, MEMORY.md, and the Cursor rules before committing or sharing them.