@hyperfixi/mcp-server 2.10.0 → 2.11.0

package/CHANGELOG.md

@@ -7,6 +7,35 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+### Changed
+
+- **MCP TypeScript SDK v2 / protocol revision 2026-07-28.** Migrated from
+  `@modelcontextprotocol/sdk` (v1) to `@modelcontextprotocol/server` `^2.0.0`.
+  The server now serves **both protocol eras** per connection via `serveStdio`:
+  the stateless `2026-07-28` revision (mandatory `server/discover`, no
+  `initialize` handshake, `resultType` stamped on every result) and the legacy
+  `initialize` dialect for current clients. Tool definitions remain plain JSON
+  Schema; the single `tools/call` dispatch path and the freshness guard are
+  unchanged.
+- **Cache hints (protocol 2026-07-28):** `tools/list`, `prompts/list`,
+  `resources/list`, and `resources/read` responses now advertise
+  `ttlMs: 3600000` / `cacheScope: "public"` — all four surfaces are static per
+  process.
+
+### Deprecated
+
+- The five sampling tools (`ask_claude`, `summarize_content`, `analyze_content`,
+  `translate_content`, `execute_llm`) rely on MCP sampling
+  (`sampling/createMessage`), which protocol revision 2026-07-28 deprecates —
+  server-initiated requests only exist on legacy-era connections whose client
+  advertises the sampling capability. They keep working there through the SDK's
+  ≥12-month deprecation window; on modern-era connections they return a clear
+  `isError` explanation. Deferred follow-up: convert them to the
+  `input_required` (MRTR) pattern before sampling's removal (~2027-07).
+  Also deferred, deliberately: per-tool `outputSchema`, tool titles/annotations,
+  and the `io.modelcontextprotocol/tasks` extension (no tool is long-running
+  enough to warrant task handles).
+
 ### Fixed
 
 - **Dispatch:** `execute_lse`, `validate_lse`, and `translate_lse` were advertised

package/README.md

@@ -1,6 +1,25 @@
 # @hyperfixi/mcp-server
 
-MCP (Model Context Protocol) server for hyperscript and multilingual DSL development. Provides **107 tools**, **9 resources**, and **9 prompts** spanning: GRAIL workflow orchestration, validation, compilation, analysis, patterns, LSP bridge, language profiles, code generation, route extraction, 9 domain DSLs, IR conversion, cross-domain dispatch, MCP sampling, AI-assisted debugging, template inventory, and the LSE round-trip pipeline.
+MCP (Model Context Protocol) server for hyperscript and multilingual DSL development. Provides **108 tools**, **9 resources**, and **9 prompts** spanning: GRAIL workflow orchestration, validation, compilation, analysis, patterns, LSP bridge, language profiles, code generation, route extraction, 9 domain DSLs, IR conversion, cross-domain dispatch, MCP sampling (opt-in), AI-assisted debugging, template inventory, and the LSE round-trip pipeline.
+
+## The agent loop
+
+The headline use case: an LLM agent emitting hyperscript instead of free-form JavaScript, with every check done by a real parser/compiler — deterministic, instant, no LLM in the checker. The server advertises this loop via MCP `instructions`, so a connected agent gets it without reading docs:
+
+1. **`validate_and_compile`** — parse the candidate into semantic IR with diagnostics; check the returned action/roles/trigger against intent.
+2. **Repair** — apply the diagnostics and re-validate. `get_code_fixes` maps coded errors to concrete fixes; `get_command_docs` / `search_patterns` show correct usage. Failed results carry a next-step hint block.
+3. **`compile_hyperscript`** — once valid, emit JavaScript (or stop at valid hyperscript for an `_="..."` attribute).
+
+To present code to a human, `translate_code` renders it in any of 24 languages via deterministic grammar transformation, `diff_behaviors` proves two snippets behaviorally equivalent, and `score_fidelity` scores a candidate against a reference — including across languages — naming exactly what is missing or hallucinated. See the repo-root [AGENTS.md](../../AGENTS.md) for a worked end-to-end example.
+
+## Protocol Support
+
+Built on the MCP TypeScript SDK v2 (`@modelcontextprotocol/server`) and serves **both protocol eras** on every stdio connection:
+
+- **`2026-07-28` (stateless):** `server/discover` answered cold, `resultType` on every result, and cache hints (`ttlMs: 3600000`, `cacheScope: "public"`) on `tools/list`, `prompts/list`, `resources/list`, and `resources/read`.
+- **Legacy (`initialize` handshake):** current clients (Claude Code, Claude Desktop) connect this way; nothing changes for them.
+
+The five sampling tools (`ask_claude`, `summarize_content`, `analyze_content`, `translate_content`, `execute_llm`) depend on MCP sampling, which revision `2026-07-28` deprecates — they work on legacy-era connections whose client advertises the sampling capability and return a clear error elsewhere. **They are also opt-in:** they invoke a generic LLM, which a connected agent can do itself, so they are hidden from `tools/list` and refuse calls unless the server is started with `LOKASCRIPT_MCP_LLM_TOOLS=1`.
 
 ## Installation
 
@@ -61,7 +80,7 @@ See **[GRAIL.md](GRAIL.md)** for full documentation, schema reference, and examp
 
 ## Available Tools
 
-The **102** tools below, plus the **5** GRAIL tools above, total **107**.
+The **103** tools below, plus the **5** GRAIL tools above, total **108**.
 
 ### Validation & Semantic Tools (8)
 
@@ -76,7 +95,7 @@ The **102** tools below, plus the **5** GRAIL tools above, total **107**.
 | `explain_in_language`  | Explain code with grammar rules, roles, and translations         |
 | `get_code_fixes`       | Get auto-fixes for specific error codes                          |
 
-### Compilation & Code Generation (6)
+### Compilation & Code Generation (7)
 
 | Tool                   | Description                                                                 |
 | ---------------------- | --------------------------------------------------------------------------- |
@@ -85,6 +104,7 @@ The **102** tools below, plus the **5** GRAIL tools above, total **107**.
 | `translate_code`       | High-fidelity translation with SVO/SOV/VSO grammar transformation           |
 | `generate_tests`       | Generate Playwright behavior tests from hyperscript                         |
 | `generate_component`   | Generate React, Vue, or Svelte component from hyperscript                   |
+| `score_fidelity`       | Score a candidate against a reference: recall/precision/role/value signals  |
 | `diff_behaviors`       | Compare two inputs for semantic equivalence                                 |
 
 ### Analysis (4)
@@ -159,7 +179,7 @@ The **102** tools below, plus the **5** GRAIL tools above, total **107**.
 | `parse_composite`   | Parse multi-line input across domains      |
 | `compile_composite` | Compile multi-line input across domains    |
 
-### MCP Sampling (5)
+### MCP Sampling (5) — opt-in via `LOKASCRIPT_MCP_LLM_TOOLS=1`
 
 | Tool                | Description                                            |
 | ------------------- | ------------------------------------------------------ |

package/dist/{chunk-P76KH42G.js → chunk-E2TYUII7.js}

@@ -206,21 +206,21 @@ async function loadDomainSchemas(domain) {
   try {
     switch (domain) {
       case "flow":
-        return (await import("@lokascript/domain-flow")).allSchemas;
+        return (await import("@lokascript/domains/flow")).allSchemas;
       case "sql":
-        return (await import("@lokascript/domain-sql")).allSchemas;
+        return (await import("@lokascript/domains/sql")).allSchemas;
       case "bdd":
-        return (await import("@lokascript/domain-bdd")).allSchemas;
+        return (await import("@lokascript/domains/bdd")).allSchemas;
       case "llm":
-        return (await import("@lokascript/domain-llm")).allSchemas;
+        return (await import("@lokascript/domains/llm")).allSchemas;
       case "jsx":
-        return (await import("@lokascript/domain-jsx")).allSchemas;
+        return (await import("@lokascript/domains/jsx")).allSchemas;
       case "todo":
-        return (await import("@lokascript/domain-todo")).allSchemas;
+        return (await import("@lokascript/domains/todo")).allSchemas;
       case "behaviorspec":
-        return (await import("@lokascript/domain-behaviorspec")).allSchemas;
+        return (await import("@lokascript/domains/behaviorspec")).allSchemas;
       case "voice":
-        return (await import("@lokascript/domain-voice")).allSchemas;
+        return (await import("@lokascript/domains/voice")).allSchemas;
       default:
         return null;
     }

package/dist/{esm-WMQN6QOB.js → esm-JVAYDN24.js}

@@ -139,7 +139,7 @@ function combine(acc, pre, values, max, maxLength, dropEmpties) {
   }
   return out;
 }
-function expandSequence(body, isAlphaSequence, max) {
+function expandSequence(body, isAlphaSequence, max, maxLength) {
   const n = body.split(/\.\./);
   const N = [];
   if (n[0] === void 0 || n[1] === void 0) {
@@ -156,6 +156,7 @@ function expandSequence(body, isAlphaSequence, max) {
     test = gte;
   }
   const pad = n.some(isPadded);
+  let length = 0;
   for (let i = x2; test(i, y) && N.length < max; i += incr) {
     let c;
     if (isAlphaSequence) {
@@ -177,7 +178,10 @@ function expandSequence(body, isAlphaSequence, max) {
         }
       }
     }
+    if (length + c.length > maxLength)
+      break;
     N.push(c);
+    length += c.length;
   }
   return N;
 }
@@ -217,7 +221,7 @@ function expand_(str, max, maxLength, isTop) {
     }
     let values;
     if (isSequence) {
-      values = expandSequence(m.body, isAlphaSequence, max);
+      values = expandSequence(m.body, isAlphaSequence, max, maxLength);
     } else {
       let n = parseCommaParts(m.body);
       if (n.length === 1 && n[0] !== void 0) {
@@ -230,9 +234,26 @@ function expand_(str, max, maxLength, isTop) {
           continue;
         }
       }
+      let dropsEmpties = dropEmpties && !m.post.length && !pre;
+      for (let d3 = 0; dropsEmpties && d3 < acc.length; d3++) {
+        if (acc[d3]) {
+          dropsEmpties = false;
+        }
+      }
       values = [];
-      for (let j2 = 0; j2 < n.length; j2++) {
-        values.push.apply(values, expand_(n[j2], max, maxLength, false));
+      let valuesLength = 0;
+      outer: for (let j2 = 0; j2 < n.length; j2++) {
+        const expanded = expand_(n[j2], max, maxLength, false);
+        for (let k2 = 0; k2 < expanded.length; k2++) {
+          const v = expanded[k2];
+          if (dropsEmpties && !v)
+            continue;
+          if (values.length >= max || valuesLength + v.length > maxLength) {
+            break outer;
+          }
+          values.push(v);
+          valuesLength += v.length;
+        }
       }
     }
     acc = combine(acc, pre, values, max, maxLength, dropEmpties && !m.post.length);

package/dist/{feedback-tools-S6VDG6F5.js → feedback-tools-6I5OQYZS.js}

@@ -2,7 +2,7 @@ import {
   feedbackTools,
   getSharedTracker,
   handleFeedbackTool
-} from "./chunk-P76KH42G.js";
+} from "./chunk-E2TYUII7.js";
 export {
   feedbackTools,
   getSharedTracker,

package/dist/index.js

@@ -2,12 +2,12 @@
 import {
   feedbackTools,
   handleFeedbackTool
-} from "./chunk-P76KH42G.js";
+} from "./chunk-E2TYUII7.js";
 
 // src/index.ts
 import { createRequire as createRequire2 } from "module";
-import { Server } from "@modelcontextprotocol/sdk/server/index.js";
-import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";
+import { Server } from "@modelcontextprotocol/server";
+import { serveStdio } from "@modelcontextprotocol/server/stdio";
 
 // src/freshness.ts
 import { createRequire } from "module";
@@ -107,16 +107,6 @@ function freshnessSummary() {
   return `watching ${watched.length} workspace dist entries`;
 }
 
-// src/index.ts
-import {
-  CallToolRequestSchema,
-  GetPromptRequestSchema,
-  ListPromptsRequestSchema,
-  ListResourcesRequestSchema,
-  ListToolsRequestSchema,
-  ReadResourceRequestSchema
-} from "@modelcontextprotocol/sdk/types.js";
-
 // src/tools/analysis.ts
 var analysisTools = [
   {
@@ -1464,7 +1454,7 @@ var validationTools = [
   },
   {
     name: "get_code_fixes",
-    description: "Get available auto-fixes for a specific error code or diagnostic. Returns LSP-compatible CodeFix suggestions that can be applied to fix common errors. Useful for LLMs to suggest corrections without running full diagnostics.",
+    description: "Get available auto-fixes for a specific error code or diagnostic \u2014 the repair half of the agent loop. When validate_and_compile or validate_hyperscript reports a coded error, this returns LSP-compatible CodeFix suggestions to apply before re-validating.",
     inputSchema: {
       type: "object",
       properties: {
@@ -5463,7 +5453,7 @@ async function getService() {
 var compilationTools = [
   {
     name: "compile_hyperscript",
-    description: 'Compile hyperscript to optimized JavaScript. Accepts natural language (code + language), explicit syntax (explicit), or LLM JSON (semantic). Common roles: patient (what to act on), destination (where to), source (where from). Use get_command_docs for per-command roles. Examples: explicit="[toggle patient:.active destination:#btn]", semantic={ action: "toggle", roles: { patient: { type: "selector", value: ".active" } } }',
+    description: 'Compile hyperscript to optimized JavaScript \u2014 the FINAL step of the agent loop (generate \u2192 validate_and_compile \u2192 repair \u2192 compile_hyperscript). Validate first: this tool reports the same failures but returns JavaScript only on success. Accepts natural language (code + language), explicit syntax (explicit), or LLM JSON (semantic). Common roles: patient (what to act on), destination (where to), source (where from). Use get_command_docs for per-command roles. Examples: explicit="[toggle patient:.active destination:#btn]", semantic={ action: "toggle", roles: { patient: { type: "selector", value: ".active" } } }',
     inputSchema: {
       type: "object",
       properties: {
@@ -5497,7 +5487,7 @@ var compilationTools = [
   },
   {
     name: "validate_and_compile",
-    description: "Parse hyperscript into semantic IR with diagnostics, without generating JavaScript. Returns action, roles, and trigger structure. Accepts natural language, explicit bracket syntax, or LLM JSON \u2014 same input formats as compile_hyperscript.",
+    description: "START HERE when generating hyperscript: parse into semantic IR with diagnostics, without generating JavaScript. Returns action, roles, and trigger structure so you can check the parse matches your intent. Accepts natural language, explicit bracket syntax, or LLM JSON \u2014 same input formats as compile_hyperscript. On failure, apply the diagnostics and re-validate (get_code_fixes maps error codes to concrete fixes; get_command_docs lists per-command roles); once valid, call compile_hyperscript for JavaScript. This validate \u2192 repair \u2192 compile loop is deterministic \u2014 no LLM in the checker.",
     inputSchema: {
       type: "object",
       properties: {
@@ -5511,7 +5501,7 @@ var compilationTools = [
   },
   {
     name: "translate_code",
-    description: "Translate hyperscript between any of 24 languages via full semantic parsing. Higher fidelity than translate_hyperscript \u2014 handles SVO/SOV/VSO grammar transformation. Preferred for production translations.",
+    description: 'Translate hyperscript between any of 24 languages via full semantic parsing \u2014 deterministic grammar transformation (SVO/SOV/VSO word order), not LLM translation. Every result carries a `verification` report (the output scored against the source via score_fidelity): verification.faithful === true is the claim "this rendering is structurally exact", so present it alongside the translation when showing code to a user for review. Higher fidelity than translate_hyperscript; preferred for production translations.',
     inputSchema: {
       type: "object",
       properties: {
@@ -5567,6 +5557,37 @@ var compilationTools = [
       }
     }
   },
+  {
+    name: "score_fidelity",
+    description: "Score a candidate hyperscript snippet against a reference for structural fidelity \u2014 the multilingual CI ratchet's deterministic scorers applied to one pair. Returns actionRecall/multisetRecall/precision/roleFidelity/valueRecall (each 0\u20131), plus the exact missing/spurious actions and lost invariant values (e.g. a silently rewritten target: toggle.destination=#panel). Sides accept any input format and may be in DIFFERENT languages, so it also verifies a translation preserved meaning. Use after editing or translating code to prove the result faithful; diff_behaviors answers identical-or-not, this answers how-faithful-and-what-drifted.",
+    inputSchema: {
+      type: "object",
+      properties: {
+        reference: {
+          type: "object",
+          description: "The trusted side (code/explicit/semantic + language)",
+          properties: {
+            code: { type: "string" },
+            explicit: { type: "string" },
+            semantic: { type: "object" },
+            language: { type: "string" }
+          }
+        },
+        candidate: {
+          type: "object",
+          description: "The side being checked (same shape as reference)",
+          properties: {
+            code: { type: "string" },
+            explicit: { type: "string" },
+            semantic: { type: "object" },
+            language: { type: "string" }
+          }
+        },
+        confidence: { type: "number", description: "Minimum confidence threshold" }
+      },
+      required: ["reference", "candidate"]
+    }
+  },
   {
     name: "diff_behaviors",
     description: "Compare two hyperscript inputs at the behavior level. Returns whether they are semantically identical, trigger diffs, and per-operation diffs. Works across languages and input formats.",
@@ -5599,6 +5620,16 @@ var compilationTools = [
     }
   }
 ];
+var REPAIR_HINT = "Next step: apply the diagnostics above and re-run validate_and_compile. If an error names a code (e.g. MISSING.ARGUMENT), get_code_fixes returns concrete fixes for it; get_command_docs lists the roles each command accepts.";
+function compileResult(result) {
+  const content = [
+    { type: "text", text: JSON.stringify(result, null, 2) }
+  ];
+  if (!result.ok) {
+    content.push({ type: "text", text: REPAIR_HINT });
+  }
+  return { content, isError: !result.ok };
+}
 async function handleCompilationTool(name, args) {
   try {
     const service = await getService();
@@ -5611,10 +5642,7 @@ async function handleCompilationTool(name, args) {
           language: args.language,
           confidence: args.confidence
         });
-        return {
-          content: [{ type: "text", text: JSON.stringify(result, null, 2) }],
-          isError: !result.ok
-        };
+        return compileResult(result);
       }
       case "validate_and_compile": {
         const result = service.validate({
@@ -5624,10 +5652,7 @@ async function handleCompilationTool(name, args) {
           language: args.language,
           confidence: args.confidence
         });
-        return {
-          content: [{ type: "text", text: JSON.stringify(result, null, 2) }],
-          isError: !result.ok
-        };
+        return compileResult(result);
       }
       case "translate_code": {
         const result = service.translate({
@@ -5670,6 +5695,17 @@ async function handleCompilationTool(name, args) {
           isError: !result.ok
         };
       }
+      case "score_fidelity": {
+        const result = service.scoreFidelity({
+          reference: args.reference,
+          candidate: args.candidate,
+          confidence: args.confidence
+        });
+        return {
+          content: [{ type: "text", text: JSON.stringify(result, null, 2) }],
+          isError: !result.ok
+        };
+      }
       case "diff_behaviors": {
         const result = service.diff({
           a: args.a,
@@ -5904,7 +5940,7 @@ async function generateServerRoutes(args) {
 }
 
 // src/tools/domain-registry-setup.ts
-import { createDomainRegistry } from "@lokascript/domain-config";
+import { createDomainRegistry } from "@lokascript/domains";
 
 // src/tools/bdd-extras.ts
 var bddScenarioParser = null;
@@ -5914,13 +5950,13 @@ async function getBDDExtras() {
   if (bddScenarioParser)
     return { parseScenario: bddScenarioParser, generator: bddGenerator, renderer: bddRenderer };
   try {
-    const mod = await import("@lokascript/domain-bdd");
+    const mod = await import("@lokascript/domains/bdd");
     bddScenarioParser = mod.parseBDDScenario;
     bddGenerator = mod.bddCodeGenerator;
     bddRenderer = mod.renderBDD ?? null;
     return { parseScenario: bddScenarioParser, generator: bddGenerator, renderer: bddRenderer };
   } catch {
-    throw new Error("@lokascript/domain-bdd not available for multi-step parsing.");
+    throw new Error("@lokascript/domains/bdd not available for multi-step parsing.");
   }
 }
 function serializeRoles(roles) {
@@ -6014,13 +6050,13 @@ async function getBehaviorSpecExtras() {
   if (behaviorSpecParser)
     return { parseSpec: behaviorSpecParser, compile: behaviorCompiler, render: behaviorRenderer };
   try {
-    const mod = await import("@lokascript/domain-behaviorspec");
+    const mod = await import("@lokascript/domains/behaviorspec");
     behaviorSpecParser = mod.parseBehaviorSpec;
     behaviorCompiler = mod.compileBehaviorSpec;
     behaviorRenderer = mod.renderBehaviorSpec;
     return { parseSpec: behaviorSpecParser, compile: behaviorCompiler, render: behaviorRenderer };
   } catch {
-    throw new Error("@lokascript/domain-behaviorspec not available for multi-line parsing.");
+    throw new Error("@lokascript/domains/behaviorspec not available for multi-line parsing.");
   }
 }
 function serializeRoles2(roles) {
@@ -6604,7 +6640,7 @@ Every value must have a role name. Do not invent commands or roles not in the sc
 var samplingTools = [
   {
     name: "ask_claude",
-    description: "Ask Claude a question with optional context. Uses MCP sampling \u2014 the client handles model selection and user approval.",
+    description: "Ask Claude a question with optional context. Uses MCP sampling \u2014 the client handles model selection and user approval. DEPRECATED: MCP sampling is deprecated as of protocol revision 2026-07-28 and works only on legacy-era connections whose client advertises the sampling capability; planned replacement is the input_required (MRTR) pattern.",
     inputSchema: {
       type: "object",
       properties: {
@@ -6631,7 +6667,7 @@ var samplingTools = [
   },
   {
     name: "summarize_content",
-    description: "Summarize text content using Claude. Uses MCP sampling \u2014 the client handles model selection and user approval.",
+    description: "Summarize text content using Claude. Uses MCP sampling \u2014 the client handles model selection and user approval. DEPRECATED: MCP sampling is deprecated as of protocol revision 2026-07-28 and works only on legacy-era connections whose client advertises the sampling capability; planned replacement is the input_required (MRTR) pattern.",
     inputSchema: {
       type: "object",
       properties: {
@@ -6658,7 +6694,7 @@ var samplingTools = [
   },
   {
     name: "analyze_content",
-    description: "Analyze text content for sentiment, entities, themes, or other qualities using Claude. Uses MCP sampling.",
+    description: "Analyze text content for sentiment, entities, themes, or other qualities using Claude. Uses MCP sampling. DEPRECATED: MCP sampling is deprecated as of protocol revision 2026-07-28 and works only on legacy-era connections whose client advertises the sampling capability; planned replacement is the input_required (MRTR) pattern.",
     inputSchema: {
       type: "object",
       properties: {
@@ -6681,7 +6717,7 @@ var samplingTools = [
   },
   {
     name: "translate_content",
-    description: "Translate text between natural languages using Claude. Uses MCP sampling \u2014 the client handles model selection and user approval.",
+    description: "Translate text between natural languages using Claude. Uses MCP sampling \u2014 the client handles model selection and user approval. DEPRECATED: MCP sampling is deprecated as of protocol revision 2026-07-28 and works only on legacy-era connections whose client advertises the sampling capability; planned replacement is the input_required (MRTR) pattern.",
     inputSchema: {
       type: "object",
       properties: {
@@ -6707,7 +6743,7 @@ var samplingTools = [
   },
   {
     name: "execute_llm",
-    description: 'Execute an LLM command written in natural language (8 languages supported). Compiles via domain-llm then invokes Claude via MCP sampling. Examples: "ask what is 2+2", "summarize #article as bullets", "\uC694\uC57D \uC774 \uBB38\uC11C\uB97C \uAE00\uBA38\uB9AC\uB85C" (Korean).',
+    description: 'Execute an LLM command written in natural language (8 languages supported). Compiles via domain-llm then invokes Claude via MCP sampling. Examples: "ask what is 2+2", "summarize #article as bullets", "\uC694\uC57D \uC774 \uBB38\uC11C\uB97C \uAE00\uBA38\uB9AC\uB85C" (Korean). DEPRECATED: MCP sampling is deprecated as of protocol revision 2026-07-28 and works only on legacy-era connections whose client advertises the sampling capability; planned replacement is the input_required (MRTR) pattern.',
     inputSchema: {
       type: "object",
       properties: {
@@ -6846,13 +6882,13 @@ ${content}`
     }
   };
 }
-async function handleSamplingTool(name, args, server2, registry2) {
+async function handleSamplingTool(name, args, server, registry2) {
   try {
     if (name === "execute_llm") {
-      return await handleExecuteLLM(args, server2, registry2);
+      return await handleExecuteLLM(args, server, registry2);
     }
     const spec = argsToSpec(name, args);
-    return await executeSpec(spec, server2);
+    return await executeSpec(spec, server);
   } catch (error) {
     const message = error instanceof Error ? error.message : String(error);
     return {
@@ -6861,13 +6897,28 @@ async function handleSamplingTool(name, args, server2, registry2) {
     };
   }
 }
-async function executeSpec(spec, server2) {
+async function executeSpec(spec, server) {
   const params = specToSamplingParams(spec);
-  const response = await server2.createMessage(params);
-  const text = response.content.type === "text" ? response.content.text : `Received non-text response of type: ${response.content.type}`;
+  let response;
+  try {
+    response = await server.createMessage(params);
+  } catch (error) {
+    const message = error instanceof Error ? error.message : String(error);
+    return {
+      content: [
+        {
+          type: "text",
+          text: `Sampling unavailable: ${message}. This tool relies on MCP sampling (deprecated in protocol revision 2026-07-28), which requires a legacy-era connection and a client that advertises the sampling capability.`
+        }
+      ],
+      isError: true
+    };
+  }
+  const block = Array.isArray(response.content) ? response.content[0] : response.content;
+  const text = block?.type === "text" ? block.text : `Received non-text response of type: ${block?.type ?? "empty"}`;
   return { content: [{ type: "text", text }] };
 }
-async function handleExecuteLLM(args, server2, registry2) {
+async function handleExecuteLLM(args, server, registry2) {
   if (!registry2) {
     return {
       content: [{ type: "text", text: "execute_llm requires domain registry (internal error)" }],
@@ -6906,12 +6957,12 @@ async function handleExecuteLLM(args, server2, registry2) {
       isError: true
     };
   }
-  return await executeSpec(spec, server2);
+  return await executeSpec(spec, server);
 }
 
 // src/tools/dispatcher.ts
 import { CrossDomainDispatcher } from "@lokascript/framework";
-import { DOMAIN_PRIORITY } from "@lokascript/domain-config";
+import { DOMAIN_PRIORITY } from "@lokascript/domains";
 var dispatcher = null;
 function getDispatcher(registry2) {
   if (!dispatcher) {
@@ -8095,7 +8146,7 @@ function createInlineExtractor() {
     const fs = await import("fs");
     const path = await import("path");
     const crypto = await import("crypto");
-    const { glob } = await import("./esm-WMQN6QOB.js");
+    const { glob } = await import("./esm-JVAYDN24.js");
     const globs = [
       "**/*.html",
       "**/*.htm",
@@ -8293,21 +8344,21 @@ async function loadDomainSchemas(domain) {
   try {
     switch (domain) {
       case "flow":
-        return (await import("@lokascript/domain-flow")).allSchemas;
+        return (await import("@lokascript/domains/flow")).allSchemas;
       case "sql":
-        return (await import("@lokascript/domain-sql")).allSchemas;
+        return (await import("@lokascript/domains/sql")).allSchemas;
       case "bdd":
-        return (await import("@lokascript/domain-bdd")).allSchemas;
+        return (await import("@lokascript/domains/bdd")).allSchemas;
       case "llm":
-        return (await import("@lokascript/domain-llm")).allSchemas;
+        return (await import("@lokascript/domains/llm")).allSchemas;
       case "jsx":
-        return (await import("@lokascript/domain-jsx")).allSchemas;
+        return (await import("@lokascript/domains/jsx")).allSchemas;
       case "todo":
-        return (await import("@lokascript/domain-todo")).allSchemas;
+        return (await import("@lokascript/domains/todo")).allSchemas;
       case "behaviorspec":
-        return (await import("@lokascript/domain-behaviorspec")).allSchemas;
+        return (await import("@lokascript/domains/behaviorspec")).allSchemas;
       case "voice":
-        return (await import("@lokascript/domain-voice")).allSchemas;
+        return (await import("@lokascript/domains/voice")).allSchemas;
       default:
         return null;
     }
@@ -8552,10 +8603,10 @@ async function handleToHyperscript(args) {
     try {
       const { CompilationService } = await import("@lokascript/compilation-service");
       const service = await CompilationService.create();
-      const compileResult = service.compile({ explicit: normalizedLse });
-      result.js = compileResult.ok ? compileResult.js : void 0;
-      if (!compileResult.ok && compileResult.diagnostics) {
-        for (const d of compileResult.diagnostics) {
+      const compileResult2 = service.compile({ explicit: normalizedLse });
+      result.js = compileResult2.ok ? compileResult2.js : void 0;
+      if (!compileResult2.ok && compileResult2.diagnostics) {
+        for (const d of compileResult2.diagnostics) {
           if (d.severity === "error") {
             diagnostics.push({
               severity: "error",
@@ -8590,8 +8641,8 @@ async function handleExecuteLse(args) {
   try {
     const { CompilationService } = await import("@lokascript/compilation-service");
     const service = await CompilationService.create();
-    const compileResult = service.compile({ explicit: normalizedLse });
-    js = compileResult.ok ? compileResult.js : void 0;
+    const compileResult2 = service.compile({ explicit: normalizedLse });
+    js = compileResult2.ok ? compileResult2.js : void 0;
   } catch {
   }
   return jsonResponse({
@@ -9498,7 +9549,7 @@ function buildCorrectedExample(attempt, diagnostics) {
 async function buildSchemaContext(domain) {
   if (!domain) return null;
   try {
-    const { loadDomainSchemas: loadDomainSchemas2 } = await import("./feedback-tools-S6VDG6F5.js");
+    const { loadDomainSchemas: loadDomainSchemas2 } = await import("./feedback-tools-6I5OQYZS.js");
     if (typeof loadDomainSchemas2 !== "function") return null;
     const schemas = await loadDomainSchemas2(domain);
     if (!schemas || schemas.length === 0) return null;
@@ -10515,150 +10566,186 @@ function readResource(uri) {
 // src/index.ts
 var registry = createDomainRegistry();
 var { version: pkgVersion } = createRequire2(import.meta.url)("../package.json");
-var server = new Server(
-  {
-    name: "@hyperfixi/mcp-server",
-    version: pkgVersion
-  },
-  {
-    capabilities: {
-      tools: {},
-      resources: {},
-      prompts: {}
+var LLM_SAMPLING_ENABLED = process.env.LOKASCRIPT_MCP_LLM_TOOLS === "1";
+var SERVER_INSTRUCTIONS = `Deterministic tooling for hyperscript (a compact DOM-behavior DSL) in 24 human languages. No tool here calls an LLM \u2014 every check is a real parser/compiler you can trust.
+
+The core loop for GENERATING hyperscript:
+1. validate_and_compile \u2014 parse your candidate into semantic IR with diagnostics; check the returned action/roles/trigger against your intent.
+2. On failure: apply the diagnostics and re-validate. get_code_fixes maps error codes to concrete fixes; get_command_docs and search_patterns show correct usage.
+3. compile_hyperscript \u2014 once valid, emit JavaScript. Or stop at valid hyperscript for an _="..." attribute.
+
+To PRESENT code to a user, translate_code renders it in any of 24 languages via deterministic grammar transformation (word-order faithful, not LLM translation); translate_to_english normalizes foreign input. diff_behaviors checks two snippets for behavioral equivalence, and score_fidelity scores a candidate against a reference (how faithful, what exactly is missing or hallucinated) \u2014 use it to prove an edit or translation preserved meaning.
+
+Prefer natural language input (code + language) for simple cases; use explicit bracket syntax ([toggle patient:.active]) or semantic JSON when you need unambiguous role control.`;
+var STATIC_SURFACE_CACHE = { ttlMs: 36e5, cacheScope: "public" };
+function buildServer() {
+  const server = new Server(
+    {
+      name: "@hyperfixi/mcp-server",
+      version: pkgVersion
+    },
+    {
+      capabilities: {
+        tools: {},
+        resources: {},
+        prompts: {}
+      },
+      instructions: SERVER_INSTRUCTIONS,
+      cacheHints: {
+        "tools/list": STATIC_SURFACE_CACHE,
+        "prompts/list": STATIC_SURFACE_CACHE,
+        "resources/list": STATIC_SURFACE_CACHE,
+        "resources/read": STATIC_SURFACE_CACHE
+      }
     }
-  }
-);
-server.setRequestHandler(ListToolsRequestSchema, async () => {
-  return {
-    tools: [
-      ...analysisTools,
-      ...patternTools,
-      ...validationTools,
-      ...lspBridgeTools,
-      ...languageDocsTools,
-      ...profileTools,
-      ...compilationTools,
-      ...routeTools,
-      ...registry.getToolDefinitions(),
-      ...samplingTools,
-      ...dispatcherTools,
-      ...irTools,
-      ...debugTools,
-      ...inventoryTools,
-      ...trainingDataTools,
-      ...feedbackTools,
-      ...lsePipelineTools,
-      ...grailTools,
-      ...lseCorrectionTools
-    ]
-  };
-});
-server.setRequestHandler(CallToolRequestSchema, async (request) => {
-  const { name, arguments: args } = request.params;
-  const stale = staleSinceStartup();
-  if (stale.length > 0) {
-    return staleToolError(stale);
-  }
-  if (name.startsWith("analyze_") || name === "explain_code" || name === "recognize_intent") {
-    return handleAnalysisTool(name, args);
-  }
-  if (name === "search_patterns" || name === "translate_hyperscript" || name === "get_pattern_stats") {
-    return handlePatternTool(name, args);
-  }
-  if (name === "validate_hyperscript" || name === "validate_schema" || name === "suggest_command" || name === "get_bundle_config" || name === "parse_multilingual" || name === "translate_to_english" || name === "explain_in_language" || name === "get_code_fixes") {
-    return handleValidationTool(name, args);
-  }
-  if (name === "get_diagnostics" || name === "get_completions" || name === "get_hover_info" || name === "get_document_symbols") {
-    return handleLspBridgeTool(name, args);
-  }
-  if (name === "get_command_docs" || name === "get_expression_docs" || name === "search_language_elements" || name === "suggest_best_practices") {
-    return handleLanguageDocsTool(name, args);
-  }
-  if (name === "get_language_profile" || name === "list_supported_languages" || name === "get_keyword_translations" || name === "get_role_markers" || name === "compare_language_profiles") {
-    return handleProfileTool(name, args);
-  }
-  if (name === "compile_hyperscript" || name === "validate_and_compile" || name === "translate_code" || name === "generate_tests" || name === "generate_component" || name === "diff_behaviors") {
-    return handleCompilationTool(name, args);
-  }
-  if (name === "detect_domain" || name === "parse_composite" || name === "compile_auto" || name === "compile_composite") {
-    return handleDispatcherTool(name, args, registry);
-  }
-  if (name === "convert_format" || name === "validate_explicit" || name === "validate_protocol" || name === "to_envelope" || name === "from_envelope") {
-    return handleIRTool(name, args);
-  }
-  if (registry.canHandle(name)) {
-    const typedArgs = args;
-    if (name.endsWith("_bdd") && isMultiStepBDD(typedArgs)) {
-      return handleBDDMultiStep(name, typedArgs);
+  );
+  server.setRequestHandler("tools/list", async () => {
+    return {
+      tools: [
+        ...analysisTools,
+        ...patternTools,
+        ...validationTools,
+        ...lspBridgeTools,
+        ...languageDocsTools,
+        ...profileTools,
+        ...compilationTools,
+        ...routeTools,
+        // Cast: the framework's MCPToolDefinition types inputSchema.properties
+        // as Record<string, unknown>, wider than the SDK's JSONValue index —
+        // the values are plain JSON Schema literals at runtime.
+        ...registry.getToolDefinitions(),
+        ...LLM_SAMPLING_ENABLED ? samplingTools : [],
+        ...dispatcherTools,
+        ...irTools,
+        ...debugTools,
+        ...inventoryTools,
+        ...trainingDataTools,
+        ...feedbackTools,
+        ...lsePipelineTools,
+        ...grailTools,
+        ...lseCorrectionTools
+      ]
+    };
+  });
+  server.setRequestHandler("tools/call", async (request) => {
+    const { name, arguments: args } = request.params;
+    const stale = staleSinceStartup();
+    if (stale.length > 0) {
+      return staleToolError(stale);
     }
-    if (name.endsWith("_behaviorspec") && isMultiLineBehaviorSpec(typedArgs)) {
-      return handleBehaviorSpecMultiLine(name, typedArgs);
+    if (name.startsWith("analyze_") || name === "explain_code" || name === "recognize_intent") {
+      return handleAnalysisTool(name, args);
     }
-    const result = await registry.handleToolCall(name, typedArgs);
-    if (result) return result;
-  }
-  if (name === "extract_routes" || name === "generate_server_routes") {
-    return handleRouteTool(name, args);
-  }
-  if (name === "ask_claude" || name === "summarize_content" || name === "analyze_content" || name === "translate_content" || name === "execute_llm") {
-    return handleSamplingTool(name, args, server, registry);
-  }
-  if (name.startsWith("debug_")) {
-    return handleDebugTool(name, args);
-  }
-  if (name === "scan_inventory" || name === "query_inventory") {
-    return handleInventoryTool(name, args);
-  }
-  if (name === "generate_training_data") {
-    return handleTrainingDataTool(name, args);
-  }
-  if (name === "lse_validate_and_feedback" || name === "lse_pattern_stats") {
-    return handleFeedbackTool(name, args);
-  }
-  if (name === "lse_from_hyperscript" || name === "lse_to_hyperscript" || name === "execute_lse" || name === "validate_lse" || name === "translate_lse") {
-    return handleLsePipelineTool(name, args);
-  }
-  if (name.startsWith("grail_")) {
-    return handleGrailTool(name, args);
-  }
-  if (name === "lse_generate_with_correction") {
-    return handleLseCorrectionTool(name, args);
-  }
-  if (name.startsWith("get_")) {
-    return handlePatternTool(name, args);
-  }
-  return {
-    content: [{ type: "text", text: `Unknown tool: ${name}` }],
-    isError: true
-  };
-});
-server.setRequestHandler(ListPromptsRequestSchema, async () => {
-  return {
-    prompts: [
-      ...getLLMPromptDefinitions(),
-      ...getDebugPromptDefinitions(),
-      ...getLSEPromptDefinitions()
-    ]
-  };
-});
-server.setRequestHandler(GetPromptRequestSchema, async (request) => {
-  const { name, arguments: promptArgs } = request.params;
-  const typedArgs = promptArgs ?? {};
-  if (isDebugPrompt(name)) {
-    return renderDebugPrompt(name, typedArgs);
-  }
-  if (isLSEPrompt(name)) {
-    return renderLSEPrompt(name, typedArgs);
-  }
-  return renderLLMPrompt(name, typedArgs);
-});
-server.setRequestHandler(ListResourcesRequestSchema, async () => {
-  return { resources: listResources() };
-});
-server.setRequestHandler(ReadResourceRequestSchema, async (request) => {
-  const { uri } = request.params;
-  return readResource(uri);
-});
+    if (name === "search_patterns" || name === "translate_hyperscript" || name === "get_pattern_stats") {
+      return handlePatternTool(name, args);
+    }
+    if (name === "validate_hyperscript" || name === "validate_schema" || name === "suggest_command" || name === "get_bundle_config" || name === "parse_multilingual" || name === "translate_to_english" || name === "explain_in_language" || name === "get_code_fixes") {
+      return handleValidationTool(name, args);
+    }
+    if (name === "get_diagnostics" || name === "get_completions" || name === "get_hover_info" || name === "get_document_symbols") {
+      return handleLspBridgeTool(name, args);
+    }
+    if (name === "get_command_docs" || name === "get_expression_docs" || name === "search_language_elements" || name === "suggest_best_practices") {
+      return handleLanguageDocsTool(name, args);
+    }
+    if (name === "get_language_profile" || name === "list_supported_languages" || name === "get_keyword_translations" || name === "get_role_markers" || name === "compare_language_profiles") {
+      return handleProfileTool(name, args);
+    }
+    if (name === "compile_hyperscript" || name === "validate_and_compile" || name === "translate_code" || name === "generate_tests" || name === "generate_component" || name === "score_fidelity" || name === "diff_behaviors") {
+      return handleCompilationTool(name, args);
+    }
+    if (name === "detect_domain" || name === "parse_composite" || name === "compile_auto" || name === "compile_composite") {
+      return handleDispatcherTool(name, args, registry);
+    }
+    if (name === "convert_format" || name === "validate_explicit" || name === "validate_protocol" || name === "to_envelope" || name === "from_envelope") {
+      return handleIRTool(name, args);
+    }
+    if (registry.canHandle(name)) {
+      const typedArgs = args;
+      if (name.endsWith("_bdd") && isMultiStepBDD(typedArgs)) {
+        return handleBDDMultiStep(name, typedArgs);
+      }
+      if (name.endsWith("_behaviorspec") && isMultiLineBehaviorSpec(typedArgs)) {
+        return handleBehaviorSpecMultiLine(name, typedArgs);
+      }
+      const result = await registry.handleToolCall(name, typedArgs);
+      if (result) return result;
+    }
+    if (name === "extract_routes" || name === "generate_server_routes") {
+      return handleRouteTool(name, args);
+    }
+    if (name === "ask_claude" || name === "summarize_content" || name === "analyze_content" || name === "translate_content" || name === "execute_llm") {
+      if (!LLM_SAMPLING_ENABLED) {
+        return {
+          content: [
+            {
+              type: "text",
+              text: `${name} is disabled: the MCP-sampling tools are opt-in (set LOKASCRIPT_MCP_LLM_TOOLS=1 in the server's environment and restart). They invoke a generic LLM, which a connected agent can do itself. For hyperscript work use validate_and_compile / compile_hyperscript; for natural-language translation of CODE, use translate_code (deterministic, not an LLM).`
+            }
+          ],
+          isError: true
+        };
+      }
+      return handleSamplingTool(name, args, server, registry);
+    }
+    if (name.startsWith("debug_")) {
+      return handleDebugTool(name, args);
+    }
+    if (name === "scan_inventory" || name === "query_inventory") {
+      return handleInventoryTool(name, args);
+    }
+    if (name === "generate_training_data") {
+      return handleTrainingDataTool(name, args);
+    }
+    if (name === "lse_validate_and_feedback" || name === "lse_pattern_stats") {
+      return handleFeedbackTool(name, args);
+    }
+    if (name === "lse_from_hyperscript" || name === "lse_to_hyperscript" || name === "execute_lse" || name === "validate_lse" || name === "translate_lse") {
+      return handleLsePipelineTool(name, args);
+    }
+    if (name.startsWith("grail_")) {
+      return handleGrailTool(name, args);
+    }
+    if (name === "lse_generate_with_correction") {
+      return handleLseCorrectionTool(name, args);
+    }
+    if (name.startsWith("get_")) {
+      return handlePatternTool(name, args);
+    }
+    return {
+      content: [{ type: "text", text: `Unknown tool: ${name}` }],
+      isError: true
+    };
+  });
+  server.setRequestHandler("prompts/list", async () => {
+    return {
+      prompts: [
+        ...getLLMPromptDefinitions(),
+        ...getDebugPromptDefinitions(),
+        ...getLSEPromptDefinitions()
+      ]
+    };
+  });
+  server.setRequestHandler("prompts/get", async (request) => {
+    const { name, arguments: promptArgs } = request.params;
+    const typedArgs = promptArgs ?? {};
+    if (isDebugPrompt(name)) {
+      return renderDebugPrompt(name, typedArgs);
+    }
+    if (isLSEPrompt(name)) {
+      return renderLSEPrompt(name, typedArgs);
+    }
+    return renderLLMPrompt(name, typedArgs);
+  });
+  server.setRequestHandler("resources/list", async () => {
+    return { resources: listResources() };
+  });
+  server.setRequestHandler("resources/read", async (request) => {
+    const { uri } = request.params;
+    return readResource(uri);
+  });
+  return server;
+}
 async function main() {
   captureFreshnessBaseline();
   const bornStale = staleAtStartup();
@@ -10670,8 +10757,7 @@ async function main() {
 ` + bornStale.map((s) => `    npm run build --prefix ${s.dir}`).join("\n")
     );
   }
-  const transport = new StdioServerTransport();
-  await server.connect(transport);
+  serveStdio(() => buildServer());
   console.error(`HyperFixi MCP server started (${freshnessSummary()})`);
 }
 main().catch((error) => {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@hyperfixi/mcp-server",
-  "version": "2.10.0",
+  "version": "2.11.0",
   "description": "Model Context Protocol server for HyperFixi hyperscript development",
   "type": "module",
   "main": "dist/index.js",
@@ -9,11 +9,11 @@
   },
   "scripts": {
     "build": "tsup src/index.ts --format esm --dts",
-    "prebuild": "../../scripts/ensure-fresh.sh ../compilation-service ../semantic ../framework ../intent ../domain-config ../domain-learn ../planner",
+    "prebuild": "../../scripts/ensure-fresh.sh ../compilation-service ../semantic ../framework ../intent ../planner",
     "start": "node dist/index.js",
     "dev": "tsx src/index.ts",
     "dev:watch": "tsx watch src/index.ts",
-    "pretest": "../../scripts/ensure-fresh.sh ../compilation-service ../semantic ../framework ../intent ../domain-config ../domain-learn ../planner",
+    "pretest": "../../scripts/ensure-fresh.sh ../compilation-service ../semantic ../framework ../intent ../planner",
     "test": "vitest",
     "typecheck": "tsc --noEmit",
     "service:setup": "echo 'Copy com.lokascript.mcp-server.plist.example to com.lokascript.mcp-server.plist and update paths, then run service:install'",
@@ -37,31 +37,22 @@
   "author": "LokaScript Contributors",
   "license": "MIT",
   "dependencies": {
-    "@hyperfixi/patterns-reference": "^2.10.0",
-    "@lokascript/compilation-service": "^2.10.0",
-    "@lokascript/domain-config": "^2.10.0",
-    "@lokascript/framework": "^2.10.0",
-    "@lokascript/planner": "^2.10.0",
-    "@lokascript/semantic": "^2.10.0",
-    "@modelcontextprotocol/sdk": "^1.25.0",
+    "@hyperfixi/patterns-reference": "^2.11.0",
+    "@lokascript/compilation-service": "^2.11.0",
+    "@lokascript/domains": "^2.11.1",
+    "@lokascript/framework": "^2.11.0",
+    "@lokascript/planner": "^2.11.0",
+    "@lokascript/semantic": "^2.11.0",
+    "@modelcontextprotocol/server": "^2.0.0",
     "yaml": "^2.8.3"
   },
   "peerDependencies": {
-    "@hyperfixi/core": "^2.10.0",
-    "@hyperfixi/patterns-reference": "^2.10.0",
-    "@hyperfixi/server-bridge": "^2.10.0",
-    "@lokascript/compilation-service": "^2.10.0",
-    "@lokascript/domain-bdd": "^2.10.0",
-    "@lokascript/domain-behaviorspec": "^2.10.0",
-    "@lokascript/domain-flow": "^2.10.0",
-    "@lokascript/domain-jsx": "^2.10.0",
-    "@lokascript/domain-learn": "^2.10.0",
-    "@lokascript/domain-llm": "^2.10.0",
-    "@lokascript/domain-sql": "^2.10.0",
-    "@lokascript/domain-todo": "^2.10.0",
-    "@lokascript/domain-voice": "^2.10.0",
-    "@lokascript/framework": "^2.10.0",
-    "@lokascript/semantic": "^2.10.0"
+    "@hyperfixi/core": "^2.11.0",
+    "@hyperfixi/patterns-reference": "^2.11.0",
+    "@hyperfixi/server-bridge": "^2.11.0",
+    "@lokascript/compilation-service": "^2.11.0",
+    "@lokascript/framework": "^2.11.0",
+    "@lokascript/semantic": "^2.11.0"
   },
   "peerDependenciesMeta": {
     "@lokascript/compilation-service": {
@@ -79,53 +70,17 @@
     "@lokascript/semantic": {
       "optional": true
     },
-    "@lokascript/domain-sql": {
-      "optional": true
-    },
-    "@lokascript/domain-bdd": {
-      "optional": true
-    },
-    "@lokascript/domain-behaviorspec": {
-      "optional": true
-    },
-    "@lokascript/domain-jsx": {
-      "optional": true
-    },
-    "@lokascript/domain-learn": {
-      "optional": true
-    },
-    "@lokascript/domain-flow": {
-      "optional": true
-    },
-    "@lokascript/domain-llm": {
-      "optional": true
-    },
-    "@lokascript/domain-todo": {
-      "optional": true
-    },
-    "@lokascript/domain-voice": {
-      "optional": true
-    },
     "@hyperfixi/server-bridge": {
       "optional": true
     }
   },
   "devDependencies": {
-    "@hyperfixi/core": "^2.10.0",
-    "@hyperfixi/patterns-reference": "^2.10.0",
-    "@hyperfixi/server-bridge": "^2.10.0",
-    "@lokascript/compilation-service": "^2.10.0",
-    "@lokascript/domain-bdd": "^2.10.0",
-    "@lokascript/domain-behaviorspec": "^2.10.0",
-    "@lokascript/domain-flow": "^2.10.0",
-    "@lokascript/domain-jsx": "^2.10.0",
-    "@lokascript/domain-learn": "^2.10.0",
-    "@lokascript/domain-llm": "^2.10.0",
-    "@lokascript/domain-sql": "^2.10.0",
-    "@lokascript/domain-todo": "^2.10.0",
-    "@lokascript/domain-voice": "^2.10.0",
-    "@lokascript/framework": "^2.10.0",
-    "@lokascript/semantic": "^2.10.0",
+    "@hyperfixi/core": "^2.11.0",
+    "@hyperfixi/patterns-reference": "^2.11.0",
+    "@hyperfixi/server-bridge": "^2.11.0",
+    "@lokascript/compilation-service": "^2.11.0",
+    "@lokascript/framework": "^2.11.0",
+    "@lokascript/semantic": "^2.11.0",
     "@types/node": "^26.1.2",
     "tsup": "^8.0.0",
     "tsx": "^4.23.1",