@getripple/mcp 1.0.4 → 1.0.5

package/CHANGELOG.md

@@ -1,5 +1,12 @@
 # @getripple/mcp Changelog
 
+## [1.0.5] - 2026-06-04
+
+### Changed
+- Refresh package README wording for npm users configuring AI agents.
+- Lead with the MCP config snippet and the before-edit / after-edit / gate / repair workflow.
+- Clarify stdio protocol behavior and local privacy posture.
+
 ## [1.0.4] - 2026-06-03
 
 ### Fixed

package/README.md

@@ -3,7 +3,14 @@
 MCP stdio server for AI agents that need Ripple's architectural context before
 editing code.
 
-Paste this into an MCP-compatible client:
+The server lets agents call Ripple directly instead of reading markdown files.
+Plan before edit. Check after edit. Catch drift. Tell the agent what to fix.
+
+## Setup
+
+Paste this into your MCP-compatible client (Claude Code, Cursor, or any MCP host).
+
+Replace the workspace path with the absolute path to your project.
 
 ```json
 {
@@ -21,38 +28,74 @@ Paste this into an MCP-compatible client:
 }
 ```
 
-The server lets agents call Ripple directly instead of scraping markdown files.
-It does not sandbox file writes by itself; MCP clients, CLI checks, CI gates, or
-human review must obey Ripple's continue/stop decisions.
+**macOS / Linux:** `/Users/yourname/projects/myapp`
+**Windows:** `C:\\Users\\yourname\\projects\\myapp`
+
+No install required. `npx` fetches and runs the server automatically.
 
-## Tools
+## Agent Workflow
 
-```txt
-ripple_get_agent_workflow
-ripple_doctor
+Call these tools in order:
+
+**1. Before editing — get context and save intent:**
+
+```
 ripple_plan_context
+```
+
+Returns the files to read first, risk level, allowed boundary, and verification
+targets. Pass `save: true` to record the intent for drift checking.
+
+**2. After editing and staging — check for drift:**
+
+```
 ripple_check_staged
-ripple_check_changed
-ripple_audit_change
+```
+
+Compares staged changes against the saved intent. Returns intent drift, boundary
+drift, policy drift, and readiness drift in a single verdict.
+
+**3. When you only need continue or stop:**
+
+```
 ripple_gate
-ripple_get_approval_status
+```
+
+Returns the compact decision the agent needs to proceed or pause for human review.
+
+**4. If drift is detected — get the repair plan:**
+
+```
 ripple_repair_intent_drift
-ripple_get_focus
-ripple_get_blast_radius
-ripple_explain_policy
-ripple_get_recent_changes
 ```
 
-Use `ripple_plan_context` before editing to save task intent and control
-boundary. Use `ripple_check_staged` or `ripple_audit_change` after editing to
-detect intent drift, boundary drift, policy drift, and readiness drift. Use
-`ripple_gate` when the agent only needs the compact continue/stop decision.
+Returns specific actions: which files to unstage, which contracts to review, and
+what to verify before widening scope.
+
+## All Tools
 
-## Local Run
+```
+ripple_get_agent_workflow      — full agent workflow guide and loop
+ripple_doctor                  — check project readiness before scanning
+ripple_plan_context            — plan before editing a target file
+ripple_check_staged            — check staged changes against saved intent
+ripple_check_changed           — check changed files against a git base ref
+ripple_audit_change            — audit a completed change for drift signals
+ripple_gate                    — compact continue/stop decision
+ripple_get_approval_status     — check whether a human gate is required
+ripple_repair_intent_drift     — get repair actions when drift is detected
+ripple_get_focus               — focused context for a single file
+ripple_get_blast_radius        — files that depend on a target file
+ripple_explain_policy          — explain the active trust boundary policy
+ripple_get_recent_changes      — recent architectural changes from history
+```
 
-After install:
+## Run Without Config
+
+After global install:
 
 ```bash
+npm install -g @getripple/mcp
 ripple-mcp --workspace /absolute/path/to/your/repo
 ```
 
@@ -62,18 +105,16 @@ Without installing:
 npx -y @getripple/mcp --workspace /absolute/path/to/your/repo
 ```
 
-For local monorepo development:
+## Protocol
 
-```bash
-npm run build:mcp
-node packages/mcp/dist/server.js --workspace /absolute/path/to/your/repo
-```
+The server communicates over stdio using newline-delimited JSON-RPC. Only
+JSON-RPC messages are written to stdout. Scan progress and logs are written to
+stderr so MCP clients can parse responses without noise.
 
-## Protocol Notes
+## Privacy
 
-The server speaks newline-delimited JSON-RPC over stdio. Only JSON-RPC messages
-are written to stdout; scan logs are redirected to stderr so MCP clients can
-parse responses safely.
+Ripple runs entirely on your machine. No account, telemetry, cloud indexing, or
+remote model call is required.
 
 ## Status
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@getripple/mcp",
-  "version": "1.0.4",
+  "version": "1.0.5",
   "description": "MCP stdio server for Ripple's local AI-agent workflow engine.",
   "license": "MIT",
   "type": "commonjs",
@@ -45,7 +45,7 @@
     "build": "tsc -p tsconfig.json"
   },
   "dependencies": {
-    "@getripple/core": "^1.0.4"
+    "@getripple/core": "^1.0.5"
   },
   "devDependencies": {
     "@types/node": "^18.0.0",