nexus-prime 3.15.1 → 3.17.0

package/README.md

@@ -68,7 +68,7 @@ Every coding agent eventually hits the same ceiling. Not model quality — infra
 
 **Nexus Prime** is a local-first MCP control plane for coding agents. It gives clients a default path that starts with bootstrap context, flows through orchestrated execution, and ends with persisted runtime truth instead of ad-hoc tool chaining.
 
-**Website:** [sir-ad.github.io/nexus-prime](https://sir-ad.github.io/nexus-prime/)<br>
+**Website:** [nexus-prime.cfd](https://nexus-prime.cfd/)<br>
 **Documentation:** [Catalog](https://sir-ad.github.io/nexus-prime/catalog.html) · [Comparison](https://sir-ad.github.io/nexus-prime/comparison.html) · [Knowledge Base](https://sir-ad.github.io/nexus-prime/knowledge-base.html) · [Integrations](https://sir-ad.github.io/nexus-prime/integrations.html) · [Architecture Diagrams](https://sir-ad.github.io/nexus-prime/architecture-diagrams.html)
 
 ## ⚡ Quick Install
@@ -129,12 +129,12 @@ Use `nexus_plan_execution` only when you explicitly want the ledger before mutat
 
 ### 2. Files read without question → Ghost Pass + Token Budgeting
 **Problem:** Agents read 25 files to answer a question that needed 3. No pre-read analysis. No decision about whether a file should even be read.
-**How:** `nexus_ghost_pass` runs a pre-read analysis pass. Source-aware token budgeting then allocates budget across repo, memory, RAG, patterns, and runtime traces — recording what was selected and what was dropped.
-**Proof:** Lifetime token telemetry, by-source allocation, per-run drilldowns survive restart.
+**How:** `nexus_ghost_pass` runs a pre-read analysis pass. Source-aware token budgeting then allocates budget across repo, memory, RAG, patterns, and runtime traces — recording what was selected and what was dropped. Tiered context loading (L0/L1/L2) progressively deepens reads based on relevance instead of loading everything at full depth. Background context compaction strips low-signal content while retaining structural elements.
+**Proof:** Lifetime token analytics with per-session breakdowns, compression ratios, and USD savings — persisted in SQLite, surviving restarts.
 
 ### 3. Memory that grows but never learns → Memory Fabric
 **Problem:** Memory accumulates without decay. Old notes outrank recent updates. No mechanism for deciding what's ephemeral vs. durable.
-**How:** The memory control plane applies fact extraction, reconciliation, quarantine, and vault projection on top of SQLite + graph base. Memories are scored, contradictions detected, and stale facts fade.
+**How:** The memory control plane applies fact extraction, reconciliation, quarantine, and vault projection on top of SQLite + graph base. Memories are scored, contradictions detected, and stale facts fade. A compaction sentinel flushes working memory to durable storage before context window compaction. Background maintenance runs on a 5-minute cycle for entropy-based expiration and deduplication.
 **Proof:** Dashboard exposes memory health, scope, trace, and shared-worker context.
 
 ### 4. RAG as prompt stuffing → Session-First RAG Gate
@@ -219,10 +219,10 @@ flowchart TD
 ## Client Setup and Runtime Contract
 
 ### Supported MCP Clients
-Nexus Prime provides automated setup for **9 coding agents**:
+Nexus Prime provides automated setup for **12 coding agents**:
 - 🔴 **Codex**
 - 🔵 **Cursor**
-- 🍊 **Claude Code**
+- 🍊 **Claude Code** (CLI + Desktop + VS Code extension)
 - 🟢 **Opencode**
 - 🌊 **Windsurf**
 - 🛡️ **Antigravity / OpenClaw**
@@ -230,7 +230,7 @@ Nexus Prime provides automated setup for **9 coding agents**:
 - 🔗 **Continue.dev**
 - ⚙️ **Cline**
 
-All agent configurations are written automatically on `npm install`. The first Nexus run inside a repo writes workspace-scoped client files. `nexus-prime setup <client>` and `nexus-prime setup all` remain the explicit refresh/fix path.
+All agent configurations are written automatically on `npm install` with XDG-aware path resolution. The first Nexus run inside a repo writes workspace-scoped client files. `nexus-prime setup <client>` and `nexus-prime setup all` remain the explicit refresh/fix path. Run `nexus-prime setup diagnose` for a dry-run validation of all client paths.
 
 ### Automated Integration
 ```bash
@@ -626,6 +626,31 @@ Inventory Snapshot: 109 skills · 64 workflows · 5 hooks · 3 automations · 7
 ## 📜 Release History
 
 <details open>
+<summary><b>v3.17.0</b> · 2026-03-20 · Agent adoption hardening, memory readability, atlas code intelligence peer</summary>
+
+- **Hard bootstrap enforcement**: Non-bootstrap MCP tool calls return a structured `blocked` response with next-step guidance instead of a soft warning agents ignored. Read-only tools remain exempt.
+- **Atlas code intelligence peer**: Auto-detects Atlas (AST parsing, semantic search, impact analysis) and writes MCP config for all clients when available.
+- **Memory readability**: Fixed text overflow on memory cards/timeline, default-filtered phantom and quarantine noise from dashboard, added phantom toggle.
+- **Imperative tool descriptions**: Rewrote MCP tool descriptions from "Preferred" to "REQUIRED" and planning nudges from "consider" to "Next: call".
+- **Streamlined CLAUDE.md**: 134-line protocol manual → 16-line imperative instruction set. Full protocol moved to `docs/nexus-protocol.md`.
+
+Full notes: [CHANGELOG.md](./CHANGELOG.md)
+</details>
+
+<details>
+<summary><b>v3.16.0</b> · 2026-03-20 · Compaction sentinel, tiered context, token analytics, peer federation, installation hardening</summary>
+
+- **Compaction Sentinel + Tiered Context**: Pre-compaction memory flush and L0/L1/L2 progressive file loading.
+- **Token Analytics**: Lifetime optimization reporting with per-session breakdowns, persisted in SQLite.
+- **Self-Learning Skills + Repo Tree**: Automatic skill derivation and cached repo tree visualization.
+- **Peer Connectors**: Auto-detect and sync with OpenClaw, Hermes, PicoClaw peers.
+- **Installation Hardening**: XDG-aware config, path validation, install logging, `nexus-prime setup diagnose`, 12 client targets.
+- **Multi-Intent Classification**: Orchestrator scores all intents, surfaces secondary when within 80% of primary.
+
+Full notes: [CHANGELOG.md](./CHANGELOG.md)
+</details>
+
+<details>
 <summary><b>v3.14.0</b> · 2026-03-16 · Ollama/HuggingFace embeddings, autonomous memory, agent ecosystem expansion</summary>
 
 - **Embedding backend chain**: Added Ollama and HuggingFace Inference API as embedding backends with automatic fallback (Ollama → HuggingFace → OpenAI → Local TF-IDF).

package/dist/agents/adapters/mcp.d.ts

@@ -29,7 +29,6 @@ export declare class MCPAdapter implements Adapter {
     private decorateToolDescription;
     debugListTools(profile?: McpToolProfile): McpToolDefinition[];
     private buildToolDefinitions;
-    private shouldAppendTelemetryFooter;
     private setupToolHandlers;
     private handleToolCall;
     scanSourceFiles(cwd: string): string[];

package/dist/agents/adapters/mcp.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"mcp.d.ts","sourceRoot":"","sources":["../../../src/agents/adapters/mcp.ts"],"names":[],"mappings":"AAYA,OAAO,EAAE,OAAO,EAAE,cAAc,EAAE,MAAM,kBAAkB,CAAC;AAC3D,OAAO,EAAE,UAAU,EAAE,MAAM,gBAAgB,CAAC;AAgE5C,KAAK,cAAc,GAAG,YAAY,GAAG,MAAM,CAAC;AAC5C,KAAK,iBAAiB,GAAG;IACrB,IAAI,EAAE,MAAM,CAAC;IACb,WAAW,EAAE,MAAM,CAAC;IACpB,WAAW,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;CACxC,CAAC;AAiKF,qBAAa,UAAW,YAAW,OAAO;IACtC,IAAI,EAAE,MAAM,CAAC;IACb,IAAI,EAAG,KAAK,CAAU;IACtB,SAAS,UAAS;IAClB,MAAM,EAAE,MAAM,EAAE,CAAM;IAEtB,OAAO,CAAC,MAAM,CAAS;IACvB,OAAO,CAAC,QAAQ,CAAC,CAAa;IAC9B,OAAO,CAAC,SAAS,CAA4C;IAC7D,OAAO,CAAC,UAAU,CAAoB;IACtC,OAAO,CAAC,OAAO,CAAC,CAAkB;IAElC,OAAO,CAAC,cAAc;IA+BtB,OAAO,CAAC,GAAG;;IAmBX,OAAO,CAAC,gBAAgB;IA4BxB,WAAW,CAAC,KAAK,EAAE,UAAU;IAI7B,OAAO,CAAC,UAAU;IAYlB,OAAO,CAAC,eAAe;IAOvB,OAAO,CAAC,cAAc;IAMtB,OAAO,CAAC,+BAA+B;IAMvC,OAAO,CAAC,uBAAuB;IAqB/B,OAAO,CAAC,uBAAuB;IAsB/B,cAAc,CAAC,OAAO,GAAE,cAAsC,GAAG,iBAAiB,EAAE;IAIpF,OAAO,CAAC,oBAAoB;IA+xB5B,OAAO,CAAC,2BAA2B;IAUnC,OAAO,CAAC,iBAAiB;YAsCX,cAAc;IAm0D5B,eAAe,CAAC,GAAG,EAAE,MAAM,GAAG,MAAM,EAAE;IAItC,OAAO,CAAC,IAAI;IAmBN,OAAO,IAAI,OAAO,CAAC,IAAI,CAAC;IAOxB,UAAU,IAAI,OAAO,CAAC,IAAI,CAAC;IAkB3B,IAAI,CAAC,OAAO,EAAE,cAAc,GAAG,OAAO,CAAC,IAAI,CAAC;IAClD,OAAO,CAAC,OAAO,EAAE,cAAc,GAAG,IAAI;CACzC"}
+{"version":3,"file":"mcp.d.ts","sourceRoot":"","sources":["../../../src/agents/adapters/mcp.ts"],"names":[],"mappings":"AAYA,OAAO,EAAE,OAAO,EAAE,cAAc,EAAE,MAAM,kBAAkB,CAAC;AAC3D,OAAO,EAAE,UAAU,EAAE,MAAM,gBAAgB,CAAC;AAiE5C,KAAK,cAAc,GAAG,YAAY,GAAG,MAAM,CAAC;AAC5C,KAAK,iBAAiB,GAAG;IACrB,IAAI,EAAE,MAAM,CAAC;IACb,WAAW,EAAE,MAAM,CAAC;IACpB,WAAW,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC;CACxC,CAAC;AAmKF,qBAAa,UAAW,YAAW,OAAO;IACtC,IAAI,EAAE,MAAM,CAAC;IACb,IAAI,EAAG,KAAK,CAAU;IACtB,SAAS,UAAS;IAClB,MAAM,EAAE,MAAM,EAAE,CAAM;IAEtB,OAAO,CAAC,MAAM,CAAS;IACvB,OAAO,CAAC,QAAQ,CAAC,CAAa;IAC9B,OAAO,CAAC,SAAS,CAA4C;IAC7D,OAAO,CAAC,UAAU,CAAoB;IACtC,OAAO,CAAC,OAAO,CAAC,CAAkB;IAElC,OAAO,CAAC,cAAc;IA+BtB,OAAO,CAAC,GAAG;;IAmBX,OAAO,CAAC,gBAAgB;IA4BxB,WAAW,CAAC,KAAK,EAAE,UAAU;IAI7B,OAAO,CAAC,UAAU;IAYlB,OAAO,CAAC,eAAe;IAOvB,OAAO,CAAC,cAAc;IAMtB,OAAO,CAAC,+BAA+B;IAMvC,OAAO,CAAC,uBAAuB;IAqB/B,OAAO,CAAC,uBAAuB;IAsB/B,cAAc,CAAC,OAAO,GAAE,cAAsC,GAAG,iBAAiB,EAAE;IAIpF,OAAO,CAAC,oBAAoB;IA8yB5B,OAAO,CAAC,iBAAiB;YAyFX,cAAc;IAy1D5B,eAAe,CAAC,GAAG,EAAE,MAAM,GAAG,MAAM,EAAE;IAItC,OAAO,CAAC,IAAI;IAmBN,OAAO,IAAI,OAAO,CAAC,IAAI,CAAC;IAOxB,UAAU,IAAI,OAAO,CAAC,IAAI,CAAC;IAmB3B,IAAI,CAAC,OAAO,EAAE,cAAc,GAAG,OAAO,CAAC,IAAI,CAAC;IAClD,OAAO,CAAC,OAAO,EAAE,cAAc,GAAG,IAAI;CACzC"}

package/dist/agents/adapters/mcp.js

@@ -20,6 +20,7 @@ import { DarwinLoop } from '../../engines/darwin-loop.js';
 import { nexusNetRelay } from '../../engines/nexusnet-relay.js';
 import { entanglementEngine, ContinuousAttentionStream, createKVBridge, nxl, OrchestratorEngine } from '../../engines/index.js';
 import { FederationEngine } from '../../engines/federation.js';
+import { TokenAnalyticsEngine } from '../../engines/token-analytics.js';
 const tokenEngine = new TokenSupremacyEngine();
 const guardrailEngine = new GuardrailEngine();
 const darwinLoop = new DarwinLoop();
@@ -61,6 +62,7 @@ const AUTONOMOUS_TOOL_ORDER = [
     'nexus_mindkit_check',
     'nexus_ghost_pass',
     'nexus_spawn_workers',
+    'nexus_token_report',
     'nexus_session_dna',
     'nexus_list_skills',
     'nexus_list_workflows',
@@ -128,6 +130,7 @@ class SessionTelemetry {
     tokensOptimized = 0;
     memoriesStored = 0;
     memoriesRecalled = 0;
+    bootstrapped = false;
     recordCall() { this.callCount++; }
     recordTokens(saved) { this.tokensOptimized += saved; }
     recordStore() { this.memoriesStored++; }
@@ -147,39 +150,39 @@ class SessionTelemetry {
         switch (event) {
             case 'recall':
                 if ((context.count ?? 0) > 3) {
-                    nudges.push('You recalled many memories — consider nexus_optimize_tokens before reading the files mentioned.');
+                    nudges.push('Next: call nexus_optimize_tokens before reading the files mentioned.');
                 }
                 if ((context.count ?? 0) === 0) {
-                    nudges.push('No memories found. This is a fresh topic — explore carefully and store key findings.');
+                    nudges.push('No memories found — fresh topic. Next: explore carefully and call nexus_store_memory with key findings.');
                 }
                 break;
             case 'optimize':
                 if ((context.fullReads ?? 0) > 3) {
-                    nudges.push('Multiple files need full reading — consider nexus_ghost_pass before modifying them.');
+                    nudges.push('Next: call nexus_ghost_pass before modifying these files.');
                 }
                 break;
             case 'ghost_pass':
                 if ((context.risks ?? 0) > 0) {
-                    nudges.push('Risks detected — strongly consider nexus_spawn_workers for parallel exploration.');
+                    nudges.push('Risks detected. Next: call nexus_spawn_workers for parallel execution.');
                 }
                 break;
             case 'store':
                 if ((context.priority ?? 0) > 0.8) {
-                    nudges.push('High-priority insight stored. Consider nexus_audit_evolution to check for recurring patterns.');
+                    nudges.push('High-priority insight stored. Next: call nexus_store_memory to persist related findings.');
                 }
                 break;
             case 'mindkit_fail':
-                nudges.push('Guardrail FAILED. Do NOT proceed. Re-scope the task or use nexus_ghost_pass for a safer approach.');
+                nudges.push('Guardrail FAILED. Do NOT proceed. Re-scope the task or call nexus_ghost_pass for a safer approach.');
                 break;
             case 'high_call_count':
                 if (this.callCount > 20) {
-                    nudges.push('You have made 20+ tool calls. Consider storing a session summary via nexus_store_memory.');
+                    nudges.push('20+ tool calls reached. Next: call nexus_store_memory with a session summary.');
                 }
                 break;
         }
         if (nudges.length === 0)
             return '';
-        return `\n<planning engine="nexus-prime">\n${nudges.map(n => `  → ${n}`).join('\n')}\n</planning>`;
+        return `\n<nexus-next-step>\n${nudges.map(n => `  → ${n}`).join('\n')}\n</nexus-next-step>`;
     }
     format(memStats) {
         const uptime = Math.round((Date.now() - this.startTime) / 1000);
@@ -335,19 +338,19 @@ export class MCPAdapter {
     }
     decorateToolDescription(name, description, profile) {
         if (name === 'nexus_session_bootstrap') {
-            return 'Preferred session-start tool for external clients. Call this first to recover memory, inspect stats, see the recommended next step, and learn whether token optimization will be applied before execution.';
+            return 'REQUIRED: Call this BEFORE any other Nexus tool or file operation. Recovers memory, context, catalog health, and execution plan. Returns recommended next step.';
         }
         if (name === 'nexus_orchestrate') {
-            return 'Preferred default for non-trivial work. Give Nexus Prime the raw request and let it choose crews, specialists, skills, workflows, hooks, automations, and token strategy automatically.';
+            return 'REQUIRED after bootstrap: Pass the user\'s raw request. Nexus handles task decomposition, skill selection, worker allocation, and execution automatically. Do NOT manually explore files instead.';
         }
         if (name === 'nexus_plan_execution') {
-            return 'Optional plan-before-run surface. Use only when you explicitly want to inspect the execution ledger before calling nexus_orchestrate.';
+            return 'Optional: Inspect the execution ledger before calling nexus_orchestrate. Skip unless you need pre-run visibility into what Nexus will choose.';
         }
         if (name === 'nexus_optimize_tokens') {
-            return 'Manual/diagnostic reading-plan tool. Usually the orchestrator applies token optimization internally; call this directly only when you need to inspect or override the reading plan.';
+            return 'Call when reading 3+ files. Returns a reading plan that saves 50-90% tokens. Follow the plan output — do not bulk-read the repo.';
         }
         if (name === 'nexus_spawn_workers') {
-            return 'Manual/diagnostic swarm surface. Prefer nexus_orchestrate unless you explicitly want low-level control over worker count, strategies, or runtime actions.';
+            return 'Call when modifying 3+ interrelated files or when nexus_ghost_pass recommends parallel execution. Prefer nexus_orchestrate for automatic worker management.';
         }
         if (MANUAL_OR_DIAGNOSTIC_TOOLS.has(name) && profile === 'full') {
             return `Advanced/manual surface. ${description}`;
@@ -500,11 +503,12 @@ export class MCPAdapter {
                 inputSchema: {
                     type: 'object',
                     properties: {
-                        task: { type: 'string', description: 'Task description' },
+                        goal: { type: 'string', description: 'What you are trying to accomplish' },
+                        task: { type: 'string', description: 'Alias for goal (deprecated, use goal)' },
                         files: { type: 'array', items: { type: 'string' }, description: 'File paths to analyze. If omitted, auto-scans src/' },
                         budget: { type: 'number', description: 'Token budget override' }
                     },
-                    required: ['task'],
+                    required: [],
                 },
             },
             // ── Mindkit ──────────────────────────────────────────────────────
@@ -598,6 +602,18 @@ export class MCPAdapter {
                 },
             },
             // ── Session DNA ──────────────────────────────────────────────────
+            {
+                name: 'nexus_token_report',
+                description: 'Get persistent token usage analytics, including lifetime tokens optimized, saved, forwarded, and cost savings across sessions.',
+                inputSchema: {
+                    type: 'object',
+                    properties: {
+                        scope: { type: 'string', enum: ['lifetime', 'recent'], description: 'Whether to get lifetime aggregate or recent sessions summary (default: lifetime)' },
+                        limit: { type: 'number', description: 'Number of recent sessions to view if scope is recent (default: 10)' }
+                    },
+                    required: [],
+                },
+            },
             {
                 name: 'nexus_session_dna',
                 description: 'Generate or load a Session DNA snapshot. Captures files accessed/modified, decisions made, skills used, and recommended next steps for perfect session handover. Use "generate" to create a snapshot of the current session, "load" to retrieve the most recent previous session\'s DNA.',
@@ -1155,15 +1171,6 @@ export class MCPAdapter {
             },
         ];
     }
-    shouldAppendTelemetryFooter(toolName) {
-        return !new Set([
-            'nexus_session_bootstrap',
-            'nexus_orchestrate',
-            'nexus_memory_stats',
-            'nexus_store_memory',
-            'nexus_ghost_pass',
-        ]).has(toolName);
-    }
     setupToolHandlers() {
         this.server.setRequestHandler(ListToolsRequestSchema, async () => {
             const profile = this.getToolProfile();
@@ -1185,14 +1192,59 @@ export class MCPAdapter {
             }
             this.telemetry.recordCall();
             this.sessionDNA.recordToolCall();
-            const result = await this.handleToolCall(request);
-            if (this.shouldAppendTelemetryFooter(String(request.params?.name ?? '')) && result.content && Array.isArray(result.content) && result.content.length > 0) {
-                const memStats = this.nexusRef.memory.getStats();
-                const footer = this.telemetry.format(memStats);
-                const last = result.content[result.content.length - 1];
-                if (last && typeof last === 'object' && 'text' in last) {
-                    last.text += footer;
-                }
+            const toolName = String(request.params?.name ?? '');
+            if (toolName === 'nexus_session_bootstrap') {
+                this.telemetry.bootstrapped = true;
+            }
+            const callId = `mcp_${Date.now()}_${Math.random().toString(36).substring(2, 9)}`;
+            const startTimeMs = Date.now();
+            nexusEventBus.emit('mcp.call.start', {
+                callId,
+                serverName: 'nexus-prime',
+                toolName,
+                args: request.params?.arguments ?? {}
+            });
+            let result;
+            try {
+                result = await this.handleToolCall(request);
+                const resultPayload = JSON.stringify(result);
+                nexusEventBus.emit('mcp.call.complete', {
+                    callId,
+                    serverName: 'nexus-prime',
+                    toolName,
+                    durationMs: Date.now() - startTimeMs,
+                    resultByteSize: Buffer.byteLength(resultPayload, 'utf8'),
+                    result
+                });
+            }
+            catch (error) {
+                nexusEventBus.emit('mcp.call.complete', {
+                    callId,
+                    serverName: 'nexus-prime',
+                    toolName,
+                    durationMs: Date.now() - startTimeMs,
+                    resultByteSize: 0,
+                    error: error.message || String(error)
+                });
+                throw error;
+            }
+            if (!this.telemetry.bootstrapped &&
+                toolName !== 'nexus_session_bootstrap' &&
+                toolName !== 'nexus_memory_stats' &&
+                !toolName.startsWith('nexus_list_') &&
+                toolName !== 'nexus_federation_status' &&
+                toolName !== 'nexus_status') {
+                return {
+                    content: [{
+                            type: 'text',
+                            text: JSON.stringify({
+                                status: 'blocked',
+                                reason: 'nexus_session_bootstrap has not been called yet',
+                                action: 'Call nexus_session_bootstrap(goal="<your task>") first, then retry this tool',
+                                hint: 'Bootstrap recovers memory and context — skipping it means working blind'
+                            }, null, 2)
+                        }]
+                };
             }
             return result;
         });
@@ -1269,6 +1321,7 @@ export class MCPAdapter {
                                     `Execution mode: ${bootstrap.recommendedExecutionMode || 'autonomous'}`,
                                     `Token optimization: ${bootstrap.tokenOptimization?.required ? 'required before broad reading' : 'not required yet'}`,
                                     `Catalog health: ${bootstrap.catalogHealth?.overall || 'unknown'} · selected ${bootstrap.artifactSelectionAudit?.selected?.length || 0}`,
+                                    `Shortlist: ${bootstrap.shortlist?.skills?.slice(0, 3).join(', ') || 'none'} (skills), ${bootstrap.shortlist?.specialists?.slice(0, 3).join(', ') || 'none'} (specialists)`,
                                     `Knowledge fabric: ${bootstrap.sourceMixRecommendation?.dominantSource || bootstrap.knowledgeFabric?.dominantSource || 'awaiting source mix'}`,
                                     `RAG: ${bootstrap.ragCandidateStatus?.attachedCollections || 0} attached · ${bootstrap.ragCandidateStatus?.retrievedChunks || 0} retrieved`,
                                     `Task graph: ${bootstrap.taskGraphPreview?.phases?.length || 0} phases · ${bootstrap.taskGraphPreview?.independentBranches || 0} branches`,
@@ -1460,15 +1513,7 @@ export class MCPAdapter {
                     content: [{
                             type: 'text',
                             text: [
-                                'Memory stored.',
-                                formatBullets([
-                                    `ID: ${id}`,
-                                    `Priority: ${priority.toFixed(2)}`,
-                                    `Tags: ${tags.join(', ') || 'none'}`,
-                                    `Content: ${content.slice(0, 140).replace(/\n+/g, ' ')}`,
-                                    autoGistNote ? autoGistNote.replace(/^\s+/, '') : 'Relay publish: skipped',
-                                ]),
-                                notification,
+                                'Memory stored. ID: ' + id,
                                 nudge,
                             ].filter(Boolean).join('\n\n'),
                         }],
@@ -1494,7 +1539,7 @@ export class MCPAdapter {
                             type: 'text',
                             text: (memories.length > 0
                                 ? `🧠 ${memories.length} memories recalled for "${query}":\n\n${memories.map((m, i) => `${i + 1}. ${m}`).join('\n\n')}`
-                                : `No memories found for "${query}". Fresh session or new topic.`) + notification + nudge,
+                                : `No memories found for "${query}". Fresh session or new topic.`) + '\n\n' + nudge,
                         }],
                 };
             }
@@ -1679,7 +1724,7 @@ export class MCPAdapter {
                 };
             }
             case 'nexus_optimize_tokens': {
-                const task = String(request.params.arguments?.task ?? '');
+                const task = String(request.params.arguments?.goal ?? request.params.arguments?.task ?? '');
                 const rawFiles = Array.isArray(request.params.arguments?.files)
                     ? request.params.arguments.files.map(String)
                     : null;
@@ -2002,6 +2047,31 @@ export class MCPAdapter {
                         }],
                 };
             }
+            case 'nexus_token_report': {
+                const scope = String(request.params.arguments?.scope ?? 'lifetime');
+                const limit = Number(request.params.arguments?.limit ?? 10);
+                const engine = new TokenAnalyticsEngine(this.nexusRef.memory);
+                if (scope === 'recent') {
+                    const history = engine.getSessionHistory(limit);
+                    return {
+                        content: [{
+                                type: 'text',
+                                text: history.length > 0
+                                    ? `📊 Recent Sessions Token Analytics:\n\n${history.map(s => `Session ID: ${s.sessionId}\n- Events: ${s.events}\n- Tokens Optimally Saved: ${s.tokensSaved.toLocaleString()} (x${s.compressionRatio.toFixed(2)} compression)\n- USD Saved: $${s.usdValueSaved.toFixed(4)}`).join('\n\n')}`
+                                    : 'No recent sessions found in token ledger.'
+                            }]
+                    };
+                }
+                else {
+                    const report = engine.getLifetimeReport();
+                    return {
+                        content: [{
+                                type: 'text',
+                                text: `📊 Lifetime Token Analytics:\n\n- Total Sessions: ${report.totalSessions}\n- Total MCP Events: ${report.totalEvents}\n- Tokens Optimally Saved: ${report.totalTokensSaved.toLocaleString()} (overall x${report.overallCompressionRatio.toFixed(2)} compression)\n- USD Saved: $${report.totalUsdSaved.toFixed(4)}`
+                            }]
+                    };
+                }
+            }
             case 'nexus_session_dna': {
                 const action = String(request.params.arguments?.action ?? 'load');
                 const sessionId = request.params.arguments?.sessionId
@@ -2013,6 +2083,7 @@ export class MCPAdapter {
                         callCount: this.telemetry.callCount ?? 0,
                         memoriesStored: this.telemetry.memoriesStored ?? 0,
                         memoriesRecalled: this.telemetry.memoriesRecalled ?? 0,
+                        tokensOptimized: this.telemetry.tokensOptimized ?? 0,
                     });
                     const dna = this.sessionDNA.flush();
                     const formatted = SessionDNAManager.format(dna);
@@ -2928,6 +2999,7 @@ export class MCPAdapter {
                 callCount: this.telemetry.callCount ?? 0,
                 memoriesStored: this.telemetry.memoriesStored ?? 0,
                 memoriesRecalled: this.telemetry.memoriesRecalled ?? 0,
+                tokensOptimized: this.telemetry.tokensOptimized ?? 0,
             });
             this.sessionDNA.flush();
             console.error('[MCP Adapter] Session DNA flushed');