diff --git a/mcp-server/CHANGELOG.md b/mcp-server/CHANGELOG.md
index f744791..4b48d62 100644
--- a/mcp-server/CHANGELOG.md
+++ b/mcp-server/CHANGELOG.md
@@ -5,6 +5,27 @@ All notable changes to `loopctl-mcp-server` are documented here.
 Format: [Keep a Changelog](https://keepachangelog.com/en/1.0.0/)
 Versioning: [Semantic Versioning](https://semver.org/spec/v2.0.0.html)
 
+## 2.2.0 — 2026-04-22 (Wiki curation tools)
+
+### Added
+
+- `knowledge_unpublish` — revert a published article back to draft. Hides it
+  from agent search/context without deleting; re-publish via
+  `knowledge_publish`. Requires `LOOPCTL_USER_KEY` (destructive, `role: :user`).
+- `knowledge_archive` — soft-delete an article (draft or published). Hidden
+  from search/context/index; row retained for audit. Requires
+  `LOOPCTL_USER_KEY`.
+- `knowledge_delete` — alias for `knowledge_archive` (DELETE verb on the REST
+  API archives under the hood). Requires `LOOPCTL_USER_KEY`.
+
+### Rationale
+
+Previously agents could create and publish articles but had no way to retract
+bad drafts via MCP — low-signal articles (session summaries, commit recaps)
+were piling up in the wiki with no cleanup path short of curl. These three
+tools close the curation loop. All three stay at `role: :user` per the
+"destructive ops above orchestrator" rule in `CLAUDE.md`.
+
 ## 2.1.0 — 2026-04-17 (Agent ergonomics)
 
 ### Added
diff --git a/mcp-server/README.md b/mcp-server/README.md
index e9a0979..c3b2ac2 100644
--- a/mcp-server/README.md
+++ b/mcp-server/README.md
@@ -66,7 +66,7 @@ Or if installed locally:
 
 Key resolution priority: `LOOPCTL_API_KEY` > tool-specific key > `LOOPCTL_ORCH_KEY`.
 
-## Tools (42)
+## Tools (45)
 
 ### Project Tools
 
@@ -145,6 +145,9 @@ Key resolution priority: `LOOPCTL_API_KEY` > tool-specific key > `LOOPCTL_ORCH_K
 |---|---|
 | `knowledge_publish` | Publish a draft article, making it visible to all agents. Required: `article_id`. |
 | `knowledge_bulk_publish` | **Requires `LOOPCTL_USER_KEY`.** Atomically publish up to 100 drafts in a single call. Required: `article_ids` (array). |
+| `knowledge_unpublish` | **Requires `LOOPCTL_USER_KEY`.** Revert a published article back to draft (hidden from search/context, not deleted). Required: `article_id`. |
+| `knowledge_archive` | **Requires `LOOPCTL_USER_KEY`.** Soft-delete an article (draft or published). Row retained for audit; hidden from all reads. Required: `article_id`. |
+| `knowledge_delete` | **Requires `LOOPCTL_USER_KEY`.** Alias for `knowledge_archive` — DELETE verb on the REST API archives under the hood. Required: `article_id`. |
 | `knowledge_drafts` | List draft (unpublished) knowledge articles with pagination. Optional: `limit` (default 20, max 20), `offset` (default 0), `project_id`. Returns `meta.total_count`. |
 | `knowledge_lint` | Run a lint check on the knowledge wiki to identify stale or low-coverage articles. Optional: `project_id`, `stale_days`, `min_coverage`, `max_per_category` (default 50, max 500). True totals returned in `summary.total_per_category`. |
 | `knowledge_export` | Export all knowledge articles as a ZIP archive. Returns a curl command for direct download (ZIP binary cannot be returned as MCP content). Optional: `project_id`. |
diff --git a/mcp-server/index.js b/mcp-server/index.js
index 4e6be67..068e7e0 100755
--- a/mcp-server/index.js
+++ b/mcp-server/index.js
@@ -652,6 +652,36 @@ async function knowledgeBulkPublish({ article_ids }) {
   return toContent(result);
 }
 
+async function knowledgeUnpublish({ article_id }) {
+  const result = await apiCall(
+    "POST",
+    `/api/v1/articles/${article_id}/unpublish`,
+    null,
+    process.env.LOOPCTL_USER_KEY
+  );
+  return toContent(result);
+}
+
+async function knowledgeArchive({ article_id }) {
+  const result = await apiCall(
+    "POST",
+    `/api/v1/articles/${article_id}/archive`,
+    null,
+    process.env.LOOPCTL_USER_KEY
+  );
+  return toContent(result);
+}
+
+async function knowledgeDelete({ article_id }) {
+  const result = await apiCall(
+    "DELETE",
+    `/api/v1/articles/${article_id}`,
+    null,
+    process.env.LOOPCTL_USER_KEY
+  );
+  return toContent(result);
+}
+
 async function knowledgeDrafts({ limit, offset, project_id }) {
   const params = new URLSearchParams();
   params.set(
@@ -1726,6 +1756,60 @@ const TOOLS = [
       required: ["article_ids"],
     },
   },
+  {
+    name: "knowledge_unpublish",
+    description:
+      "Revert a published article back to draft state. The article stops being visible " +
+      "in agent search/context but is not deleted — re-publish with knowledge_publish. " +
+      "REQUIRES LOOPCTL_USER_KEY (user role — orchestrator role is NOT sufficient for " +
+      "this destructive operation).",
+    inputSchema: {
+      type: "object",
+      properties: {
+        article_id: {
+          type: "string",
+          description: "The UUID of the published article to unpublish.",
+        },
+      },
+      required: ["article_id"],
+    },
+  },
+  {
+    name: "knowledge_archive",
+    description:
+      "Archive an article (soft delete). The article is hidden from search, context, " +
+      "and the index but the row is retained for audit/history. Works for drafts and " +
+      "published articles. REQUIRES LOOPCTL_USER_KEY (user role — orchestrator role is " +
+      "NOT sufficient for this destructive operation).",
+    inputSchema: {
+      type: "object",
+      properties: {
+        article_id: {
+          type: "string",
+          description: "The UUID of the article to archive.",
+        },
+      },
+      required: ["article_id"],
+    },
+  },
+  {
+    name: "knowledge_delete",
+    description:
+      "Delete an article. Under the hood this performs the same soft-delete (archive) " +
+      "as knowledge_archive — use whichever name is clearer at the call site. The row " +
+      "is retained for audit; there is no hard delete. REQUIRES LOOPCTL_USER_KEY (user " +
+      "role — orchestrator role is NOT sufficient for this destructive operation).",
+    inputSchema: {
+      type: "object",
+      properties: {
+        article_id: {
+          type: "string",
+          description: "The UUID of the article to delete.",
+        },
+      },
+      required: ["article_id"],
+    },
+  },
   {
     name: "knowledge_drafts",
     description:
@@ -2233,6 +2317,15 @@ server.setRequestHandler(CallToolRequestSchema, async (request) => {
     case "knowledge_bulk_publish":
       return await knowledgeBulkPublish(args);
 
+    case "knowledge_unpublish":
+      return await knowledgeUnpublish(args);
+
+    case "knowledge_archive":
+      return await knowledgeArchive(args);
+
+    case "knowledge_delete":
+      return await knowledgeDelete(args);
+
     case "knowledge_drafts":
       return await knowledgeDrafts(args);
 
diff --git a/mcp-server/package.json b/mcp-server/package.json
index 094a2a6..32fbd05 100644
--- a/mcp-server/package.json
+++ b/mcp-server/package.json
@@ -1,6 +1,6 @@
 {
   "name": "loopctl-mcp-server",
-  "version": "2.1.0",
+  "version": "2.2.0",
   "description": "MCP server for loopctl — structural trust for AI development loops",
   "type": "module",
   "main": "index.js",