@plur-ai/mcp 0.2.1 → 0.2.2

package/README.md

@@ -1,18 +1,10 @@
 # @plur-ai/mcp
 
-MCP server for PLUR — gives any AI agent persistent memory via the Model Context Protocol.
+Give your AI agent persistent memory. One line in your MCP config — corrections, preferences, and conventions persist across sessions. No workflow changes, no cloud, no API costs for search.
 
-```bash
-npx @plur-ai/mcp
-```
-
-Or install globally:
+Part of [PLUR](https://plur.ai) — where **Haiku with memory outperforms Opus without it** at 10x less cost.
 
-```bash
-npm install -g @plur-ai/mcp
-```
-
-## Setup
+## Setup (30 seconds)
 
 ### Claude Code
 
@@ -38,55 +30,93 @@ Add to `.cursor/mcp.json`:
 }
 ```
 
-### Custom storage path
+### Windsurf / any MCP client
 
-```json
-{
-  "mcpServers": {
-    "plur": {
-      "command": "npx",
-      "args": ["-y", "@plur-ai/mcp"],
-      "env": { "PLUR_PATH": "/path/to/storage" }
-    }
-  }
-}
+Same pattern — point it at `npx -y @plur-ai/mcp`.
+
+That's it. Your agent now has memory. Use your tools as usual — corrections accumulate automatically.
+
+## What happens
+
+```
+You correct your agent  →  engram created       →  YAML on your disk
+Next session starts     →  relevant ones injected →  agent remembers
+You rate the result     →  engram strengthens    →  quality improves
 ```
 
+Knowledge is stored as **engrams** — small assertions that strengthen with use and decay when irrelevant. Search is fully local (BM25 + embeddings), so memory recall costs nothing and works offline. **86.7% on LongMemEval** — on par with cloud memory services.
+
 ## Tools
 
-| Tool | Description |
+Your agent gets these tools automatically:
+
+| Tool | What it does |
 |------|-------------|
 | `plur.learn` | Store a memory — correction, preference, convention, or decision |
 | `plur.recall` | Keyword search (BM25, instant) |
-| `plur.recall.hybrid` | **Best default** — BM25 + embeddings merged via RRF. No API calls. |
-| `plur.inject` | Select engrams for current task within token budget |
-| `plur.feedback` | Rate an engram — trains injection relevance over time |
+| `plur.recall.hybrid` | **Best default** — BM25 + embeddings merged via RRF. Zero cost. |
+| `plur.inject` | Load relevant memories for the current task |
+| `plur.feedback` | Rate a memory — trains relevance over time |
 | `plur.forget` | Retire a memory (history preserved) |
-| `plur.capture` | Record a session event to the episodic timeline |
-| `plur.timeline` | Query past episodes by time, agent, or search |
-| `plur.ingest` | Extract engram candidates from text |
-| `plur.sync` | Git-based sync across machines |
-| `plur.sync.status` | Check sync state (initialized, remote, dirty, ahead/behind) |
-| `plur.packs.install` | Install an engram pack |
+| `plur.sync` | Sync memory across machines via git |
+| `plur.sync.status` | Check sync state |
+| `plur.capture` | Record a session event |
+| `plur.timeline` | Query session history |
+| `plur.ingest` | Extract learnings from text |
+| `plur.packs.install` | Install a shareable memory pack |
 | `plur.packs.list` | List installed packs |
-| `plur.status` | System health — engram count, episodes, packs, storage root |
+| `plur.status` | System health |
 
-## How agents use it
+## Sync across machines
 
-**Session start:** Call `plur.inject` with the task description to load relevant memory.
+Your agent can sync memory to any git remote:
+
+```
+Agent: plur.sync({ remote: "git@github.com:you/plur-memory.git" })
+→ "Initialized and pushed."
+
+# On another machine, same remote:
+Agent: plur.sync()
+→ "Synced. Pulled 12 remote commits."
+```
+
+Works with GitHub, GitLab, Gitea, any git host. Your data, your repo.
+
+## Configuration
+
+Custom storage path:
+
+```json
+{
+  "mcpServers": {
+    "plur": {
+      "command": "npx",
+      "args": ["-y", "@plur-ai/mcp"],
+      "env": { "PLUR_PATH": "/path/to/storage" }
+    }
+  }
+}
+```
 
-**When corrected:** Call `plur.learn` to store the correction.
+Default: `~/.plur/`. Everything is plain YAML — open it, read it, edit it.
 
-**Session end:** Call `plur.capture` to record what happened.
+## Benchmark
 
-**Rate results:** Call `plur.feedback` on injected engrams to improve future quality.
+| Metric | Score |
+|--------|-------|
+| LongMemEval overall | **86.7%** |
+| A/B win rate vs no memory | 89% |
+| House rules accuracy | 100% |
 
-**Cross-device:** Call `plur.sync` with a git remote URL to sync memory across machines.
+[Full methodology →](https://plur.ai/benchmark.html)
 
-## Update notifications
+## Related packages
 
-The server checks npm on startup and logs a warning if a newer version is available. No overhead during operation — the check runs once, asynchronously.
+| Package | For |
+|---------|-----|
+| [`@plur-ai/core`](https://www.npmjs.com/package/@plur-ai/core) | Engine — use directly in custom agent frameworks |
+| [`@plur-ai/claw`](https://www.npmjs.com/package/@plur-ai/claw) | OpenClaw — automatic memory without MCP |
 
 ## License
 
-Apache-2.0
+Apache-2.0 · [GitHub](https://github.com/plur-ai/plur) · [plur.ai](https://plur.ai)

package/dist/index.js

@@ -348,7 +348,7 @@ function getToolDefinitions() {
 }
 
 // src/server.ts
-var VERSION = "0.2.1";
+var VERSION = "0.2.2";
 async function createServer(plur) {
   const instance = plur ?? new Plur();
   const tools = getToolDefinitions();

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@plur-ai/mcp",
-  "version": "0.2.1",
+  "version": "0.2.2",
   "type": "module",
   "bin": {
     "plur-mcp": "dist/index.js"
@@ -12,7 +12,7 @@
   "dependencies": {
     "@modelcontextprotocol/sdk": "^1.12.0",
     "zod": "^3.23.0",
-    "@plur-ai/core": "0.2.1"
+    "@plur-ai/core": "0.2.2"
   },
   "devDependencies": {
     "@types/node": "^25.5.0"