Staging Simulated 6 MCP tools

Staging Engine

Classify diffs into atomic Commit-Scopes before an agent commits.

How you use it

Before every commit: classify → pick one Commit-Scope → stash unrelated paths → staging_enforce → commit_checkpoint. Never invent a scope to match a mixed index.

Token savings

Stops mega-commits that force the next agent to re-read half the repo. Atomic SHAs mean smaller diffs, cheaper reviews, and fewer “what changed?” rediscovery loops.

Consistency

Exactly one staging category per commit. Message Commit-Scope must match staged paths — no Gate-override for mixed batches.

In action

Staging
Incoming diff

Agent touched auth + docs + a new util in one tree.

Classify
single-change
Batch blocked

Interactive demo

Drag a diff onto a Commit-Scope bucket, or select a diff and click a scope. Simulated staging only.

Pending diffs

Commit-Scopes

1 Diff detected
2 Classify scope
3 Scope locked
4 Staging enforce
5 Commit-ready

0/4 classified · scrub playback to walk the pipeline

Staging pipeline Idle

Simulated behavior only — does not execute proprietary AGS engine code.

Diagram

Example usage

// MCP: staging_enforce
{
  "commitScope": "function-create",
  "message": "status/checkpoint(auth): add refresh\n\nCommit-Scope: function-create"
}

// MCP: staging_classify
{ "includeUnstaged": true }
Docs →
  • staging_classify
  • staging_validate
  • staging_diff
  • staging_enforce
  • staging_split
  • staging_stash