@agentmemory/agentmemory 0.7.7 → 0.7.9

package/README.md

@@ -3,8 +3,8 @@
 </p>
 
 <p align="center">
-  <strong>Persistent memory for AI coding agents.</strong><br/>
-  Powered by <a href="https://iii.dev">iii-engine</a>.
+  <strong>Your coding agent remembers everything. No more re-explaining.</strong><br/>
+  Persistent memory for Claude Code, Cursor, Gemini CLI, OpenCode, and any MCP client.
 </p>
 
 <p align="center">
@@ -26,9 +26,11 @@
 
 ---
 
-Every AI coding agent has the same blind spot. Session ends, memory vanishes. You re-explain architecture. You re-discover bugs. You re-teach preferences. Built-in memory files like CLAUDE.md and .cursorrules are 200-line sticky notes that overflow and go stale. agentmemory replaces that with a searchable, versioned, cross-agent database — 43 MCP tools, triple-stream retrieval (BM25 + vector + knowledge graph), 4-tier memory consolidation, provenance-tracked citations, and cascading staleness so retired facts never pollute your context again. One instance serves Claude Code, Cursor, Codex, Windsurf, and any MCP client simultaneously. 627 tests. Zero external DB dependencies.
+You explain the same architecture every session. You re-discover the same bugs. You re-teach the same preferences. Built-in memory (CLAUDE.md, .cursorrules) caps out at 200 lines and goes stale. agentmemory fixes this — it silently captures what your agent does, compresses it into searchable memory, and injects the right context when the next session starts. One command. Works across agents.
 
-The result is measurable. On 240 real observations across 30 sessions, agentmemory hits 64% Recall@10 and perfect MRR while using 92% fewer tokens than dumping everything into context. When an agent searches "database performance optimization," it finds the N+1 fix you made three weeks ago — something keyword grep literally cannot do. Memories version automatically, supersede each other, propagate staleness to related graph nodes, and sync across agent instances via P2P mesh. Your agents stop repeating mistakes. Your context stays clean. Your sessions start fast.
+**What changes:** Session 1 you set up JWT auth. Session 2 you ask for rate limiting — the agent already knows your auth uses jose middleware in `src/middleware/auth.ts`, your tests cover token validation, and you chose jose over jsonwebtoken for Edge compatibility. No re-explaining. No copy-pasting. The agent just *knows*.
+
+**95.2% retrieval accuracy** on [LongMemEval](https://arxiv.org/abs/2410.10813) (ICLR 2025). 43 MCP tools. 12 hooks. Real-time viewer. Works with Claude Code, Cursor, Gemini CLI, OpenCode, and any MCP client. 646 tests. Zero external DB dependencies.
 
 ```bash
 npx @agentmemory/agentmemory   # installs iii-engine if missing, starts everything
@@ -38,7 +40,7 @@ npx @agentmemory/agentmemory   # installs iii-engine if missing, starts everythi
 
 ## Why agentmemory
 
-AI coding agents forget everything between sessions. You explain the same architecture, re-discover the same patterns, and re-learn the same preferences every time. agentmemory fixes that.
+Every coding agent forgets everything when the session ends. You waste the first 5 minutes of every session re-explaining your stack, your conventions, your recent decisions. agentmemory runs in the background and eliminates that entirely.
 
 ```
 Session 1: "Add auth to the API"
@@ -47,15 +49,14 @@ Session 1: "Add auth to the API"
   Session ends -> observations compressed into structured memory
 
 Session 2: "Now add rate limiting"
-  agentmemory injects context from Session 1:
+  Agent already knows:
     - Auth uses JWT middleware in src/middleware/auth.ts
     - Tests in test/auth.test.ts cover token validation
-    - Decision: chose jose over jsonwebtoken for Edge compatibility
-  Agent starts with full project awareness
+    - You chose jose over jsonwebtoken for Edge compatibility
+    - The rate limit discussion from last week's debugging session
+  Zero re-explaining. Starts working immediately.
 ```
 
-No manual notes. No copy-pasting. The agent just *knows*.
-
 ### What it gives you
 
 | Capability | What it does |
@@ -97,14 +98,25 @@ agentmemory is the searchable database behind the sticky notes.
 | Multi-agent coordination | Impossible | Leases, signals, actions, routines |
 | Cross-agent sync | No | P2P mesh (7 scopes: memories, actions, semantic, procedural, relations, graph) |
 | Memory trust | No verification | Citation chain back to source observations with confidence scores |
-| Semantic search | No (keyword grep) | Yes (Recall@10: 64% vs 56% for grep) |
+| Semantic search | No (keyword grep) | Yes (95.2% R@5 on LongMemEval-S) |
 | Memory lifecycle | Manual pruning | Ebbinghaus decay + tiered eviction |
 | Knowledge graph | No | Entity extraction + temporal versioning |
 | Observability | Read files manually | Real-time viewer on :3113 |
 
 ### Benchmarks (measured, not projected)
 
-Evaluated on 240 real-world coding observations across 30 sessions with 20 labeled queries:
+#### LongMemEval-S (ICLR 2025, 500 questions)
+
+Evaluated on [LongMemEval-S](https://arxiv.org/abs/2410.10813), an academic benchmark with 500 questions across ~48 sessions per question (~115K tokens). Same dataset and metric (`recall_any@K`) used by other memory systems.
+
+| System | R@5 | R@10 | NDCG@10 | MRR |
+|---|---|---|---|---|
+| **agentmemory BM25+Vector** | **95.2%** | **98.6%** | **87.9%** | **88.2%** |
+| agentmemory BM25-only | 86.2% | 94.6% | 73.0% | 71.5% |
+
+These are retrieval recall scores (not end-to-end QA accuracy). Embedding model: `all-MiniLM-L6-v2` (local, no API key).
+
+#### Internal benchmark (240 observations, 20 queries)
 
 | System | Recall@10 | NDCG@10 | MRR | Tokens/query |
 |---|---|---|---|---|
@@ -112,9 +124,9 @@ Evaluated on 240 real-world coding observations across 30 sessions with 20 label
 | agentmemory BM25 (stemmed + synonyms) | 55.9% | 82.7% | 95.5% | 1,571 |
 | agentmemory + Xenova embeddings | **64.1%** | **94.9%** | **100.0%** | **1,571** |
 
-With real embeddings, agentmemory finds "N+1 query fix" when you search "database performance optimization" — something keyword matching literally cannot do.
+agentmemory finds "N+1 query fix" when you search "database performance optimization" — something keyword matching literally cannot do.
 
-Full benchmark reports: [`benchmark/QUALITY.md`](benchmark/QUALITY.md), [`benchmark/SCALE.md`](benchmark/SCALE.md), [`benchmark/REAL-EMBEDDINGS.md`](benchmark/REAL-EMBEDDINGS.md)
+Full benchmark reports: [`benchmark/LONGMEMEVAL.md`](benchmark/LONGMEMEVAL.md), [`benchmark/QUALITY.md`](benchmark/QUALITY.md), [`benchmark/SCALE.md`](benchmark/SCALE.md), [`benchmark/REAL-EMBEDDINGS.md`](benchmark/REAL-EMBEDDINGS.md)
 
 ## Supported Agents
 
@@ -135,9 +147,10 @@ Any agent that connects to MCP servers can use agentmemory's 43 tools, 6 resourc
 
 | Agent | How to connect |
 |---|---|
+| **Cursor** | Add MCP server in settings or `~/.cursor/mcp.json` |
 | **Claude Desktop** | Add to `claude_desktop_config.json` MCP servers |
-| **Cursor** | Add MCP server in settings |
-| **Windsurf** | MCP server configuration |
+| **Gemini CLI** | `gemini mcp add agentmemory -- npx agentmemory-mcp` |
+| **OpenCode** | Add to `.opencode/config.json` MCP servers |
 | **Cline / Continue** | MCP server configuration |
 | **Any MCP client** | Point to `http://localhost:3111/agentmemory/mcp/*` |
 
@@ -160,13 +173,30 @@ GET  /agentmemory/profile       # Get project intelligence
 |---|---|
 | Claude Code user | Plugin install (hooks + MCP + skills) |
 | Building a custom agent with Claude SDK | AgentSDKProvider (zero config) |
-| Using Cursor, Windsurf, or any MCP client | MCP server (41 tools + 6 resources + 3 prompts) |
+| Using Cursor, Gemini CLI, OpenCode, or any MCP client | MCP server (43 tools + 6 resources + 3 prompts) |
 | Building your own agent framework | REST API (103 endpoints) |
 | Sharing memory across multiple agents | All agents point to the same iii-engine instance |
 
 ## Quick Start
 
-### 1. Install the Plugin (Claude Code)
+### 1. Start agentmemory
+
+```bash
+npx @agentmemory/agentmemory
+```
+
+This auto-installs iii-engine if missing, starts it, and runs the worker. One command.
+
+Or from source:
+
+```bash
+git clone https://github.com/rohitg00/agentmemory.git && cd agentmemory
+npm install && npm run build && npm start
+```
+
+### 2. Connect your agent
+
+**Claude Code (plugin — hooks + MCP + skills):**
 
 ```bash
 /plugin marketplace add rohitg00/agentmemory
@@ -175,35 +205,66 @@ GET  /agentmemory/profile       # Get project intelligence
 
 All 12 hooks, 4 skills, and MCP server are registered automatically.
 
-### 2. Start agentmemory
+**Cursor / Claude Desktop / Cline / any MCP client:**
+
+Add to your MCP config (e.g. `~/.cursor/mcp.json`, `claude_desktop_config.json`):
+
+```json
+{
+  "mcpServers": {
+    "agentmemory": {
+      "command": "npx",
+      "args": ["agentmemory-mcp"]
+    }
+  }
+}
+```
+
+**Gemini CLI:**
 
 ```bash
-npx @agentmemory/agentmemory
+gemini mcp add agentmemory -- npx agentmemory-mcp
 ```
 
-This auto-installs iii-engine if missing, starts it, and runs the worker. One command.
+**OpenCode:**
 
-Or from source:
+Add to `.opencode/config.json`:
+
+```json
+{
+  "mcpServers": {
+    "agentmemory": {
+      "command": "npx",
+      "args": ["agentmemory-mcp"]
+    }
+  }
+}
+```
+
+**REST API (any agent, any language):**
 
 ```bash
-git clone https://github.com/rohitg00/agentmemory.git && cd agentmemory
-npm install && npm run build && npm start
+curl -X POST http://localhost:3111/agentmemory/remember \
+  -H "Content-Type: application/json" \
+  -d '{"content": "Always use jose for JWT on Edge", "type": "preference"}'
+
+curl -X POST http://localhost:3111/agentmemory/smart-search \
+  -H "Content-Type: application/json" \
+  -d '{"query": "JWT authentication"}'
 ```
 
 ### 3. Verify
 
 ```bash
 curl http://localhost:3111/agentmemory/health
-
-# Real-time viewer (auto-starts on port 3113)
-open http://localhost:3113
+open http://localhost:3113   # Real-time viewer
 ```
 
 ```json
 {
   "status": "healthy",
   "service": "agentmemory",
-  "version": "0.7.4",
+  "version": "0.7.7",
   "health": {
     "memory": { "heapUsed": 42000000, "heapTotal": 67000000 },
     "cpu": { "percent": 2.1 },
@@ -424,7 +485,7 @@ Collects every 30 seconds: heap usage, CPU percentage (delta sampling), event lo
 
 ## MCP Server
 
-### Tools (38)
+### Tools (43)
 
 | Tool | Description |
 |------|-------------|
@@ -488,7 +549,7 @@ Collects every 30 seconds: heap usage, CPU percentage (delta sampling), event lo
 
 ### Standalone MCP Server
 
-Run agentmemory as a standalone MCP server for any MCP-compatible agent (Cursor, Codex, Gemini CLI, Windsurf):
+Run agentmemory as a standalone MCP server for any MCP-compatible agent (Cursor, Gemini CLI, OpenCode, Claude Desktop, Cline):
 
 ```bash
 npx agentmemory-mcp
@@ -616,7 +677,7 @@ ANTHROPIC_API_KEY=sk-ant-...
 # Obsidian Export (v0.7.0)
 # OBSIDIAN_AUTO_EXPORT=false
 
-# MCP Tool Visibility (v0.7.0) — "core" (7 tools) or "all" (41 tools)
+# MCP Tool Visibility (v0.7.0) — "core" (7 tools) or "all" (43 tools)
 # AGENTMEMORY_TOOLS=core
 
 # Team Memory (v0.5.0)
@@ -724,9 +785,9 @@ agentmemory is built on iii-engine's three primitives:
 | Prometheus / Grafana | iii OTEL + built-in health monitor |
 | Redis (circuit breaker) | In-process circuit breaker + fallback chain |
 
-**113 source files. ~20,000 LOC. 627 tests. Zero external DB dependencies.**
+**118 source files. ~21,800 LOC. 646 tests. Zero external DB dependencies.**
 
-### Functions (89 mem:: functions)
+### Functions (123 mem:: functions)
 
 | Category | Functions | Purpose |
 |----------|-----------|---------|
@@ -808,7 +869,7 @@ agentmemory is built on iii-engine's three primitives:
 ```bash
 npm run dev               # Hot reload
 npm run build             # Production build (~425KB)
-npm test                  # Unit tests (627 tests, ~1.5s)
+npm test                  # Unit tests (646 tests, ~1.7s)
 npm run test:integration  # API tests (requires running services)
 ```
 

package/dist/cli.mjs

@@ -160,12 +160,12 @@ async function main() {
 	p.intro("agentmemory");
 	if (skipEngine) {
 		p.log.info("Skipping engine check (--no-engine)");
-		await import("./src-C_TC9frp.mjs");
+		await import("./src-DNbB7fd7.mjs");
 		return;
 	}
 	if (await isEngineRunning()) {
 		p.log.success("iii-engine is running");
-		await import("./src-C_TC9frp.mjs");
+		await import("./src-DNbB7fd7.mjs");
 		return;
 	}
 	if (!await startEngine()) {
@@ -193,7 +193,7 @@ async function main() {
 		process.exit(1);
 	}
 	s.stop("iii-engine is ready");
-	await import("./src-C_TC9frp.mjs");
+	await import("./src-DNbB7fd7.mjs");
 }
 main().catch((err) => {
 	p.log.error(err instanceof Error ? err.message : String(err));

package/dist/index.mjs

@@ -3805,7 +3805,7 @@ function registerAutoForgetFunction(sdk, kv) {
 
 //#endregion
 //#region src/version.ts
-const VERSION = "0.7.7";
+const VERSION = "0.7.9";
 
 //#endregion
 //#region src/functions/export-import.ts
@@ -3909,7 +3909,8 @@ function registerExportImportFunction(sdk, kv) {
 			"0.7.4",
 			"0.7.5",
 			"0.7.6",
-			"0.7.7"
+			"0.7.7",
+			"0.7.9"
 		]).has(importData.version)) return {
 			success: false,
 			error: `Unsupported export version: ${importData.version}`