Tony363 · Tony363 · Dec 16, 2025 · Dec 16, 2025 · Dec 16, 2025 · Dec 16, 2025
diff --git a/.codex-os/product/analysis.md b/.codex-os/product/analysis.md
@@ -23,7 +23,7 @@ repo_ref: main@d151479
 # 2. Product Context
 - Decisions log (2025-10-17, 2025-10-25, 2025-10-26) documents MCP consolidation, UnifiedStore adoption, and stricter `requires_evidence` handling, but no mission/roadmap docs exist locally.
 - Intended users are Claude Code operators who need verifiable diffs with offline guarantees; success metrics would center on executed vs plan-only rates, loop iterations required, and quality scores.
-- Constraints: offline mode by default, limited MCP roster (Sequential, Zen, Deepwiki), reliance on git/git tests for evidence, and absence of networked verification.
+- Constraints: offline mode by default, limited MCP roster (Sequential, PAL, Deepwiki), reliance on git/git tests for evidence, and absence of networked verification.
 
 # 3. Architecture Overview
 - **Entry:** `/sc:` commands parsed via registry metadata (`SuperClaude/Commands/registry.py`) and executed through `CommandExecutor`.

diff --git a/.codex-os/product/decisions.md b/.codex-os/product/decisions.md
@@ -12,7 +12,7 @@
 
 ## 2025-10-25 – MCP Simplification & UnifiedStore
 - **Context:** Maintaining six local MCP stubs created redundant documentation, extra configuration, and brittle command dependencies. Serena’s JSON persistence also diverged from the desired SQLite-backed storage.
-- **Decision:** Retire Context7, Magic, MorphLLM, Playwright, and Serena MCP integrations. Introduce the `UnifiedStore` SQLite backend with a migration helper, and update commands/modes/docs to rely on Sequential, Zen, and Deepwiki only.
+- **Decision:** Retire Context7, Magic, MorphLLM, Playwright, and Serena MCP integrations. Introduce the `UnifiedStore` SQLite backend with a migration helper, and update commands/modes/docs to rely on Sequential, PAL, and Deepwiki only.
 - **Consequences:** MCP registry, installer components, and command playbooks now reference a minimal, actively-supported toolset. Session persistence flows through UnifiedStore, and automated tests cover the new storage path (`tests/test_worktree_state.py`).
 
 ## 2025-10-25 – Auto-Stub Hygiene & Agent Telemetry

diff --git a/.codex-os/product/fast-codex-execution-plan.md b/.codex-os/product/fast-codex-execution-plan.md
@@ -3,7 +3,7 @@
 ## 0. Context
 - Suggestion: extend `/sc:implement` with a streamlined `--fast-codex` flag that keeps the existing command executor but runs a simplified persona and validation path.
 - Motivation: reduce friction for routine edits while preserving guardrails such as evidence capture, MCP activation, and telemetry hooks.
-- Constraints: offline-first operation, limited MCP roster (Zen + sequential stubs), `requires_evidence` guardrail must remain intact, and docs/tests must stay aligned with `.codex-os` standards.
+- Constraints: offline-first operation, limited MCP roster (PAL + sequential stubs), `requires_evidence` guardrail must remain intact, and docs/tests must stay aligned with `.codex-os` standards.
 - Assumptions: command metadata remains YAML front matter, executor logic is centralized in `SuperClaude/Commands/executor.py`, and consensus policies live in `SuperClaude/Config/consensus_policies.yaml`.
 
 ## 1. Goals & Success Metrics

diff --git a/.codex-os/product/hallucination-mitigation-plan.md b/.codex-os/product/hallucination-mitigation-plan.md
@@ -3,7 +3,7 @@
 ## 0. Context
 - Source analysis: `.codex-os/product/analysis.md` (2025-10-27).
 - Target outcomes: Reduce plan-only completions, ensure consensus uses real ensembles, surface hallucination regressions within CI/observability, and ground agent outputs with verifiable evidence.
-- Constraints: Offline-first runtime, limited MCP roster (Sequential, Zen, Deepwiki), rely on git evidence and local tests.
+- Constraints: Offline-first runtime, limited MCP roster (Sequential, PAL, Deepwiki), rely on git evidence and local tests.
 
 ## 1. Goals & Success Metrics
 - **G1 – Reliable consensus:** All `requires_evidence` commands succeed only after multi-model agreement.

diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -18,7 +18,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   setup CLI against regressions when file lists change.
 
 ### Changed
-- Removed Context7, Magic, MorphLLM, Playwright, Serena, and Deepwiki MCP integrations; streamlined registry to Sequential and Zen.
+- Removed Context7, Magic, MorphLLM, Playwright, Serena, and Deepwiki MCP integrations; streamlined registry to Sequential and PAL.
 - Updated commands, modes, docs, and agents to reference UnifiedStore and remaining MCP servers.
 - Simplified installer components and MCP documentation to match current server lineup.
 - Core and Modes installer components now persist the selected profile, expanded file manifests,

diff --git a/Docs/Developer-Guide/technical-architecture.md b/Docs/Developer-Guide/technical-architecture.md
@@ -46,9 +46,9 @@ contributors can map features to code quickly.
 - `ModelRouterFacade` resolves provider clients (OpenAI, Anthropic, Google,
   X.AI). When credentials are missing it returns a structured error so callers
   can decide whether to retry or short-circuit.
-- `ZenIntegration` (MCP) now delegates to `ModelRouterFacade.run_consensus`,
-  re-packaging the payload into a simple dataclass. The integration therefore
-  shares the same provider availability rules as the core executor.
+- MCP functionality (consensus, code review, etc.) is now accessed via Claude
+  Code's native tools (`mcp__pal__*`, `mcp__rube__*`) instead of custom wrappers.
+  No Python integration code is needed.
 - Tests that need deterministic behaviour register in-memory executors directly
   on `ConsensusBuilder`.
 

diff --git a/Docs/Developer-Guide/testing-debugging.md b/Docs/Developer-Guide/testing-debugging.md
@@ -106,7 +106,7 @@ directory (command artefacts, metrics JSONL, etc.). Inspecting those artefacts i
 the fastest way to understand failures because they reflect the actual executor
 output.
 
-- When API keys are available, `--zen-review` (GPT-5) now becomes the primary
-  `QualityScorer` evaluator during `--loop`. Tests that stub zen responses should
-  assert on `QualityDimension.ZEN_REVIEW` metrics or the GPT-provided
+- When API keys are available, `--pal-review` (GPT-5) now becomes the primary
+  `QualityScorer` evaluator during `--loop`. Tests that stub pal responses should
+  assert on `QualityDimension.PAL_REVIEW` metrics or the GPT-provided
   `improvements` list as part of their expectations.
diff --git a/Docs/Reference/troubleshooting.md b/Docs/Reference/troubleshooting.md
@@ -14,7 +14,7 @@ Follow these steps when commands do not behave as expected.
 ## MCP Problems
 
 - Run `python -m SuperClaude.MCP --list` to view enabled servers or
-  `python -m SuperClaude.MCP --describe zen` for details on a specific adapter.
+  `python -m SuperClaude.MCP --describe pal` for details on a specific adapter.
 - Set `SC_NETWORK_MODE=debug` to capture detailed HTTP traces.
 - For Rube automation, confirm `SC_RUBE_API_KEY` is present or switch to dry-run
   mode.

@@ -29,6 +29,27 @@ and the artefacts they generate.
   guardrails.
 - `--delegate <agent>`: Pin execution to a specific agent.
 - `--loop`: Allow iterative agent hand-offs until success criteria are met.
+- `--pal-review`: Request PAL code review after changes are applied.
+
+## PAL Review Integration
+
+When the `--pal-review` flag is used, the executor signals that a code review
+should be performed. The review is handled by Claude invoking the native MCP
+tool `mcp__pal__codereview` after the Python executor completes.
+
+**Workflow:**
+1. Command executes with `--pal-review` flag
+2. Executor applies changes and sets `pal_review_signal` in results
+3. Claude reads the signal and invokes `mcp__pal__codereview`
+4. Review feedback is incorporated before task completion
+
+**Example:**
+```
+/sc:implement refactor auth module --pal-review
+```
+
+This pattern enables the full Plan→Act→Review→Refine loop while respecting
+the architecture where MCP tools are invoked at the conversation level.
 
 See the individual command modules under `SuperClaude/Commands/` for extended
 options and YAML front matter describing required context.
diff --git a/Docs/User-Guide/mcp-servers.md b/Docs/User-Guide/mcp-servers.md
@@ -11,13 +11,13 @@ have been removed.
 - The CLI honours `SC_NETWORK_MODE`. Set it to `offline` to skip network calls,
   `online` for full access, or `debug` to log request payloads.
 - Server-specific environment variables:
-  - `SC_ZEN_OFFLINE=1` to force Zen into offline mode (requires manual executor
+  - `SC_PAL_OFFLINE=1` to force PAL into offline mode (requires manual executor
     registration).
   - `SC_RUBE_API_KEY` for Rube automation calls and LinkUp web searches.
 
-## 2. Zen Integration (Consensus)
+## 2. PAL Integration (Consensus)
 
-- The Zen adapter now uses `ModelRouterFacade.run_consensus`, which means it
+- The PAL adapter now uses `ModelRouterFacade.run_consensus`, which means it
   requires the same provider executors as the rest of the framework.
 - When no executors are available the integration raises `RuntimeError` and the
   command fails fast—there is no heuristic fallback.

diff --git a/Docs/memory-optimization-plan.md b/Docs/memory-optimization-plan.md
@@ -19,7 +19,7 @@ Each milestone will be checked off as the work lands.
 - **Core component** installs every Markdown file in `SuperClaude/Core/` (e.g., `RULES_*`, `OPERATIONS.md`, `AGENTS*.md`, `BUSINESS_*`). The heaviest contributors are `OPERATIONS.md`, `WORKFLOWS.md`, `AGENTS_EXTENDED.md`, and the business panel guides (>2 k tokens each).
 - **Agents component** copies all top-level persona `.md` files in `SuperClaude/Agents/`, including the extended catalogues and business assistants.
 - **Modes component** imports all `MODE_*.md` files even though most sessions only require `MODE_Normal.md` and `MODE_Task_Management.md`.
-- **MCP docs** default to installing `MCP_Zen.md`, `MCP_Rube.md`, and `MCP_LinkUp.md`. Their footprint is modest, but they still count toward the memory bundle.
+- **MCP docs** default to installing `MCP_Pal.md`, `MCP_Rube.md`, and `MCP_LinkUp.md`. Their footprint is modest, but they still count toward the memory bundle.
 
 ### Profile Definitions (current state)
 

diff --git a/Docs/real_integrations_plan.md b/Docs/real_integrations_plan.md
@@ -12,14 +12,11 @@ current implementation, remaining risks, and optional follow-up experiments.
 - **Nice-to-have:** add a smoke test that runs the bootstrap flow inside the
   benchmark harness’s virtualenv job.
 
-## Rube MCP Live Mode (`SuperClaude/MCP/rube_integration.py`)
-- **Current state:** live HTTP calls include retry with backoff, structured
-  error propagation, and telemetry tags (`rube_mcp`). Dry-run remains available
-  via `SC_RUBE_MODE=dry-run`.
-- **What to monitor next:** record circuit-breaker metrics once we add provider
-  rate limits and expose a health summary in `.superclaude_metrics`.
-- **Nice-to-have:** ship contract fixtures for partner sandboxes so CI can run
-  smoke requests when credentials exist.
+## Rube MCP (Native Tools)
+- **Current state:** MCP functionality is now accessed via Claude Code's native
+  tools (`mcp__rube__*`). No custom HTTP wrapper is needed.
+- **What to monitor next:** usage patterns of native MCP tools in command flows.
+- **Nice-to-have:** add command-level telemetry for MCP tool invocations.
 
 ## Token Accounting (`SuperClaude/Monitoring/performance_monitor.py`)
 - **Current state:** every provider invocation updates cumulative counters and

diff --git a/README.md b/README.md
@@ -35,7 +35,7 @@ SuperClaude is a sophisticated AI orchestration framework that enhances Claude C
 - **100+ Specialized Agents**: From backend architects to security auditors
 - **Behavioral Modes**: Normal, Task Management, Token Efficiency, Orchestration
 - **Quality Validation**: Multi-stage pipelines with syntax, security, and performance checks
-- **MCP Server Integration**: Rube (web/app automation), Zen (consensus), LinkUp (web search)
+- **MCP Server Integration**: Rube (web/app automation), PAL (consensus), LinkUp (web search)
 
 ```mermaid
 graph TB
@@ -58,7 +58,7 @@ graph TB
 
         Executor --> MCP[MCP Integrations]
         MCP --> Rube[Rube MCP]
-        MCP --> Zen[Zen MCP]
+        MCP --> PAL[PAL MCP]
 
         Executor --> Quality[Quality Pipeline]
         Quality --> Validation[Validation Stages]
@@ -107,7 +107,7 @@ flowchart TB
         API_G[Google]
         API_X[xAI]
         MCP_R[Rube MCP]
-        MCP_Z[Zen MCP]
+        MCP_P[PAL MCP]
     end
 
     subgraph Storage["Storage & State"]
@@ -438,7 +438,7 @@ classDiagram
         +behavior_mode: str
         +think_level: int
         +loop_enabled: bool
-        +zen_review_enabled: bool
+        +pal_review_enabled: bool
     }
 
     class CommandResult {
@@ -495,7 +495,7 @@ graph TB
         Executor[Command Executor]
 
         Executor --> RubeInt[Rube Integration]
-        Executor --> ZenInt[Zen Integration]
+        Executor --> PALInt[PAL Integration]
 
         subgraph "Rube MCP"
             RubeInt --> Tools[500+ App Tools]
@@ -506,68 +506,56 @@ graph TB
             Tools --> LinkUp[LinkUp Search]
         end
 
-        subgraph "Zen MCP"
-            ZenInt --> Consensus[Consensus Engine]
+        subgraph "PAL MCP"
+            PALInt --> Consensus[Consensus Engine]
             Consensus --> Review[Code Review]
             Review --> GPT5[GPT-5 Analysis]
         end
     end
 ```
 
-#### Rube MCP
+#### Rube MCP (Native)
 
-Rube MCP connects 500+ apps for seamless cross-app automation.
+Rube MCP connects 500+ apps for seamless cross-app automation via Claude Code's native tools.
 
-```python
-from SuperClaude.MCP.rube_integration import RubeIntegration
-
-# Initialize Rube
-rube = RubeIntegration({
-    "endpoint": "https://rube.app/mcp",
-    "api_key": os.getenv("SC_RUBE_API_KEY"),
-    "enabled": True
-})
-
-# Web search via LinkUp
-result = await rube.linkup_search(
-    query="latest React 19 features",
-    depth="deep",
-    output_type="sourcedAnswer"
-)
-
-# Batch searches
-results = await rube.linkup_batch_search(
-    queries=["Python 3.13 features", "TypeScript 5.6 changes"],
-    max_concurrent=4
-)
+```
+# Web search via LinkUp - use mcp__rube__RUBE_MULTI_EXECUTE_TOOL
+Use mcp__rube__RUBE_MULTI_EXECUTE_TOOL with:
+  tools: [{
+    "tool_slug": "LINKUP_SEARCH",
+    "arguments": {
+      "query": "latest React 19 features",
+      "depth": "deep",
+      "output_type": "sourcedAnswer"
+    }
+  }]
+  session_id: "<from RUBE_SEARCH_TOOLS>"
+  memory: {}
 ```
 
-#### Zen MCP
-
-Zen MCP provides local consensus orchestration and code review.
+#### PAL MCP (Native) - Formerly "Zen"
 
-```python
-from SuperClaude.MCP.zen_integration import ZenIntegration, ConsensusType
-
-# Initialize Zen
-zen = ZenIntegration()
-await zen.initialize_session()
-
-# Run consensus
-result = await zen.consensus(
-    prompt="Evaluate this architectural decision",
-    models=[ModelConfig("gpt-5"), ModelConfig("claude-opus-4.5")],
-    vote=ConsensusType.weighted,
-    thinking=ThinkingMode.high
-)
+PAL MCP provides consensus orchestration and code review via Claude Code's native tools.
 
-# Code review
-review = await zen.review_code(
-    diff=git_diff_content,
-    files=["src/auth.py", "src/api.py"],
-    model="gpt-5",
-    max_issues=10
-)
+```
+# Code review - use mcp__pal__codereview
+Use mcp__pal__codereview with:
+  step: "Review authentication module for security issues"
+  step_number: 1
+  total_steps: 2
+  next_step_required: true
+  findings: "Initial security scan..."
+  relevant_files: ["/path/to/auth.py"]
+  model: "gpt-5.2"
+
+# Multi-model consensus - use mcp__pal__consensus
+Use mcp__pal__consensus with:
+  step: "Evaluate: Should we use REST or GraphQL?"
+  step_number: 1
+  total_steps: 3
+  next_step_required: true
+  findings: "Analyzing tradeoffs..."
+  models: [{"model": "gpt-5.2", "stance": "for"}, {"model": "gemini-3-pro", "stance": "against"}]
 ```
 
 ---
@@ -839,7 +827,7 @@ SuperClaude integrates with Claude Code via `CLAUDE.md` configuration files:
 
 # MCP Documentation
 @SuperClaude/Core/MCP_Rube.md
-@SuperClaude/Core/MCP_Zen.md
+@SuperClaude/Core/MCP_Pal.md
 ```
 
 ### Project-Level Configuration
@@ -929,7 +917,7 @@ pie title Agent Distribution
 | `--think` | 1-5 | Thinking depth level |
 | `--loop` | iterations | Enable quality iteration |
 | `--consensus` | majority/unanimous | Consensus strategy |
-| `--zen-review` | true/false | Enable GPT-5 code review |
+| `--pal-review` | true/false | Enable GPT-5 code review |
 
 ### Quality Flags
 
@@ -1052,8 +1040,10 @@ SuperClaude/
 │   │   └── AGENTS.md           # Agent guidelines
 │   │
 │   ├── MCP/
-│   │   ├── rube_integration.py # Rube MCP client
-│   │   └── zen_integration.py  # Zen consensus
+│   │   ├── __init__.py        # Native MCP tools reference
+│   │   ├── MCP_Rube.md        # Rube MCP documentation
+│   │   ├── MCP_Pal.md         # PAL MCP documentation
+│   │   └── MCP_LinkUp.md      # LinkUp search documentation
 │   │
 │   ├── ModelRouter/
 │   │   ├── router.py           # Request routing

diff --git a/SECURITY.md b/SECURITY.md
@@ -278,7 +278,7 @@ For organizations requiring extended security support:
 
 **Individual Server Security:**
 
-**Zen**: Consensus orchestration with deterministic validation paths
+**PAL**: Consensus orchestration with deterministic validation paths
 **Rube**: Automation hub with scoped OAuth delegation and dry-run defaults
 **Browser**: Local Chromium bridge with sandboxed automation and credential gating
 **UnifiedStore**: Session persistence with secure local storage and access controls