diff --git a/packages/loopover-mcp/bin/loopover-mcp.js b/packages/loopover-mcp/bin/loopover-mcp.js
index 93c03b6bb9..ec0ce3cee1 100755
--- a/packages/loopover-mcp/bin/loopover-mcp.js
+++ b/packages/loopover-mcp/bin/loopover-mcp.js
@@ -30,6 +30,8 @@ import { buildSlopAssessment, SLOP_RUBRIC_MARKDOWN } from "@loopover/engine/sign
 import { buildTestEvidenceReport } from "@loopover/engine/signals/test-evidence";
 // #6754: the same pure evaluator the remote MCP tool + /v1/loop/evaluate-escalation both call.
 import { evaluateEscalation } from "@loopover/engine";
+// #6752: the same pure composer the remote MCP tool + /v1/loop/results-payload both call.
+import { buildResultsPayload } from "@loopover/engine";
 import { z } from "zod";
 import { buildBranchAnalysisPayload, collectLocalDiff, collectLocalBranchMetadata, probeLocalScorer, referenceScorePreviewExample, resolveScorePreviewCommand, resolveWorkspaceCwd, sanitizeLocalScorerStatus, setupGuidanceForLocalScorer, isTestFile } from "../lib/local-branch.js";
 import { formatTable } from "../lib/format-table.js";
@@ -545,6 +547,19 @@ const evaluateEscalationShape = {
   killRequested: z.boolean().optional(),
 };
 
+// #6752: mirrors buildResultsPayloadShape in src/mcp/server.ts exactly, so the local tool, the remote tool, and
+// the REST route all accept an identical payload.
+const resultsPayloadShape = {
+  repoFullName: z.string().min(1),
+  prNumber: z.number().int().nullable().optional(),
+  title: z.string(),
+  changedFiles: z
+    .array(z.object({ path: z.string(), additions: z.number().int().optional(), deletions: z.number().int().optional() }))
+    .max(5000)
+    .optional(),
+  status: z.enum(["open", "merged", "closed"]).optional(),
+};
+
 // #6749: mirrors checkTestEvidenceShape in src/mcp/server.ts VERBATIM (same bounds, same optionality).
 const checkTestEvidenceShape = {
   changedPaths: z.array(z.string().min(1).max(400)).max(2000),
@@ -890,6 +905,12 @@ const STDIO_TOOL_DESCRIPTORS = [
     description:
       "Decide whether a rented loop needs a human, and what action to take, from an already-computed run outcome, health tier, and operator/customer signals — the deterministic support/escalation-path logic. Source-free; returns shouldEscalate + action (none/notify/human_review/stop) + severity + reasons. It decides; the caller wires the action. Computed in-process; no API round-trip.",
   },
+  {
+    name: "loopover_build_results_payload",
+    category: "agent",
+    description:
+      "Package a completed loop iteration into the customer-facing result (#4801): a PR link, a plain-language summary, and a bounded diff preview, from already-computed iteration metadata. Deterministic and source-free — it formats the result, it does not fetch, open, or deliver anything. Computed in-process; no API round-trip.",
+  },
   {
     name: "loopover_check_issue_slop",
     category: "review",
@@ -1487,6 +1508,18 @@ registerStdioTool(
   (input) => toolResult("LoopOver escalation decision.", evaluateEscalation(input)),
 );
 
+registerStdioTool(
+  "loopover_build_results_payload",
+  {
+    description: stdioToolDescription("loopover_build_results_payload"),
+    inputSchema: resultsPayloadShape,
+  },
+  // Computed in-process from @loopover/engine (#6752) — the same pure buildResultsPayload the remote server
+  // (src/mcp/server.ts) and the /v1/loop/results-payload route both call, so all three surfaces return an
+  // identical payload for identical input, and results composition works fully offline.
+  (input) => toolResult("LoopOver loop results payload.", buildResultsPayload(input)),
+);
+
 registerStdioTool(
   "loopover_check_issue_slop",
   {
diff --git a/src/api/routes.ts b/src/api/routes.ts
index c6bc539bab..9dd2530839 100644
--- a/src/api/routes.ts
+++ b/src/api/routes.ts
@@ -200,6 +200,7 @@ import { runIssueRagRetrieval, validateIssueRagInput, type IssueRagInput } from
 import { buildBoundaryTestGenerationFinding, buildBoundaryTestGenerationSpec } from "../signals/boundary-test-generation";
 import { buildTestEvidenceReport } from "../signals/test-evidence";
 import { evaluateEscalation } from "../loop-escalation";
+import { buildResultsPayload } from "../results-payload";
 import { loadPrAiReviewFindings } from "../mcp/pr-ai-review-findings";
 import {
   buildMcpCompatibilityMetadata,
@@ -490,6 +491,19 @@ const evaluateEscalationSchema = z.object({
   killRequested: z.boolean().optional(),
 });
 
+// #6752: mirrors buildResultsPayloadShape in src/mcp/server.ts VERBATIM (same bounds, same optionality) so the
+// REST surface can never accept an input the MCP tool would reject, or vice versa.
+const resultsPayloadSchema = z.object({
+  repoFullName: z.string().min(1),
+  prNumber: z.number().int().nullable().optional(),
+  title: z.string(),
+  changedFiles: z
+    .array(z.object({ path: z.string(), additions: z.number().int().optional(), deletions: z.number().int().optional() }))
+    .max(5000)
+    .optional(),
+  status: z.enum(["open", "merged", "closed"]).optional(),
+});
+
 // #6749: mirrors checkTestEvidenceShape in src/mcp/server.ts VERBATIM (same bounds, same optionality) so the
 // REST surface can never accept an input the MCP tool would reject, or vice versa.
 const testEvidenceSchema = z.object({
@@ -3293,6 +3307,18 @@ export function createApp() {
     return c.json(evaluateEscalation(parsed.data));
   });
 
+  // #6752: REST mirror of the loopover_build_results_payload MCP tool, bringing it to the same REST/CLI parity
+  // its same-tier sibling loopover_check_slop_risk (/v1/lint/slop-risk) already has. Both are pure, source-free
+  // composers over caller-supplied, already-computed iteration metadata, so this route delegates to the same
+  // buildResultsPayload the tool calls and adds no logic of its own -- it formats the result, it does not fetch,
+  // open, or deliver anything.
+  app.post("/v1/loop/results-payload", async (c) => {
+    const body = await c.req.json().catch(() => null);
+    const parsed = resultsPayloadSchema.safeParse(body);
+    if (!parsed.success) return c.json({ error: "invalid_results_payload_request", issues: parsed.error.issues }, 400);
+    return c.json(buildResultsPayload(parsed.data));
+  });
+
   app.post("/v1/lint/issue-slop", async (c) => {
     const body = await c.req.json().catch(() => null);
     const parsed = issueSlopSchema.safeParse(body);
diff --git a/test/unit/mcp-cli-results-payload-tool.test.ts b/test/unit/mcp-cli-results-payload-tool.test.ts
new file mode 100644
index 0000000000..2c5a6e7661
--- /dev/null
+++ b/test/unit/mcp-cli-results-payload-tool.test.ts
@@ -0,0 +1,88 @@
+import { Client } from "@modelcontextprotocol/sdk/client/index.js";
+import { StdioClientTransport } from "@modelcontextprotocol/sdk/client/stdio.js";
+import { mkdtempSync, rmSync } from "node:fs";
+import { tmpdir } from "node:os";
+import { join } from "node:path";
+import { afterEach, beforeEach, describe, expect, it } from "vitest";
+import { buildResultsPayload, type IterationResult } from "../../src/results-payload";
+
+// #6752: the local mirror of loopover_build_results_payload. Like its same-tier sibling loopover_check_slop_risk,
+// it composes IN-PROCESS from @loopover/engine — no API round-trip — so results composition works fully offline.
+// The point of these tests is cross-surface PARITY: the stdio tool must return exactly what the pure
+// buildResultsPayload returns for identical input (the same function /v1/loop/results-payload delegates to).
+const bin = join(process.cwd(), "packages/loopover-mcp/bin/loopover-mcp.js");
+
+let client: Client;
+let transport: StdioClientTransport;
+let configDir: string;
+
+beforeEach(async () => {
+  configDir = mkdtempSync(join(tmpdir(), "loopover-results-payload-"));
+  transport = new StdioClientTransport({
+    command: "node",
+    args: [bin, "--stdio"],
+    // Pure + in-process: a black-holed API URL proves no round-trip happens.
+    env: { ...process.env, LOOPOVER_CONFIG_DIR: configDir, LOOPOVER_TOKEN: "session-token", LOOPOVER_API_URL: "http://127.0.0.1:1", LOOPOVER_API_TIMEOUT_MS: "1000" },
+  });
+  client = new Client({ name: "results-payload-test", version: "0.0.1" });
+  await client.connect(transport);
+});
+
+afterEach(async () => {
+  await client?.close().catch(() => undefined);
+  if (configDir) rmSync(configDir, { recursive: true, force: true });
+});
+
+describe("loopover_build_results_payload stdio mirror (#6752)", () => {
+  it("registers the tool alongside its same-tier check_slop_risk sibling", async () => {
+    const names = new Set((await client.listTools()).tools.map((t) => t.name));
+    expect(names).toContain("loopover_build_results_payload");
+    expect(names).toContain("loopover_check_slop_risk");
+  });
+
+  it("matches the pure composer for every input shape — offline, with no API reachable", async () => {
+    const cases: IterationResult[] = [
+      { repoFullName: "acme/widgets", prNumber: 42, title: "Opened", status: "open" },
+      { repoFullName: "acme/widgets", prNumber: 42, title: "Merged", status: "merged" },
+      { repoFullName: "acme/widgets", prNumber: 42, title: "Closed", status: "closed" },
+      { repoFullName: "acme/widgets", prNumber: 7, title: "Status omitted defaults to open" },
+      { repoFullName: "acme/widgets", title: "prNumber absent entirely" },
+      { repoFullName: "acme/widgets", prNumber: null, title: "prNumber null" },
+      { repoFullName: "acme/widgets", prNumber: 9, title: "Empty changed set", changedFiles: [] },
+      { repoFullName: "acme/widgets", prNumber: 9, title: "Counts omitted", changedFiles: [{ path: "README.md" }] },
+      {
+        repoFullName: "acme/widgets",
+        prNumber: 9,
+        title: "Over the preview cap",
+        changedFiles: Array.from({ length: 12 }, (_, i) => ({ path: `src/f${i}.ts`, additions: i, deletions: 1 })),
+      },
+    ];
+    for (const args of cases) {
+      const result = await client.callTool({ name: "loopover_build_results_payload", arguments: args });
+      expect(result.isError, JSON.stringify(args)).toBeFalsy();
+      // PARITY: identical to what the REST route returns, because both call this same function.
+      expect((result as { structuredContent?: unknown }).structuredContent, JSON.stringify(args)).toEqual(
+        JSON.parse(JSON.stringify(buildResultsPayload(args))),
+      );
+    }
+  });
+
+  it("rejects invalid input (zod input-schema validation)", async () => {
+    for (const args of [
+      {},
+      { title: "missing repoFullName" },
+      { repoFullName: "", title: "empty repoFullName" },
+      { repoFullName: "acme/widgets" },
+      { repoFullName: "acme/widgets", title: 7 },
+      { repoFullName: "acme/widgets", title: "bad status", status: "reopened" },
+      { repoFullName: "acme/widgets", title: "bad prNumber", prNumber: 1.5 },
+      { repoFullName: "acme/widgets", title: "bad changedFiles", changedFiles: [{ additions: 1 }] },
+    ]) {
+      const rejected = await client.callTool({ name: "loopover_build_results_payload", arguments: args }).then(
+        (r) => Boolean(r.isError),
+        () => true,
+      );
+      expect(rejected, `${JSON.stringify(args)} should be rejected`).toBe(true);
+    }
+  });
+});
diff --git a/test/unit/mcp-tool-rename-aliases.test.ts b/test/unit/mcp-tool-rename-aliases.test.ts
index 79c718bbf0..dce95ced24 100644
--- a/test/unit/mcp-tool-rename-aliases.test.ts
+++ b/test/unit/mcp-tool-rename-aliases.test.ts
@@ -10,6 +10,7 @@
 // (#6621 registered the loopover_get_eligibility_plan REST/CLI mirror, taking the count from 61 to 62.)
 // (#6615 registered the loopover_close_pr write-tool — 9th of the 9 buildXSpec builders — taking the count from 62 to 63.)
 // (#6732 registered the loopover_monitor_open_prs CLI mirror, taking the count from 63 to 64.)
+// (#6752 registered the loopover_build_results_payload CLI mirror, taking the count from 67 to 68.)
 import { Client } from "@modelcontextprotocol/sdk/client/index.js";
 import { StdioClientTransport } from "@modelcontextprotocol/sdk/client/stdio.js";
 import { mkdtempSync, rmSync } from "node:fs";
@@ -53,14 +54,14 @@ describe("MCP legacy alias retirement (#4777) — discovery invariants", () => {
   });
   afterEach(disconnect);
 
-  it("lists exactly 67 loopover_ tools and zero gittensory_-prefixed aliases", async () => {
+  it("lists exactly 68 loopover_ tools and zero gittensory_-prefixed aliases", async () => {
     const { tools } = await client.listTools();
     const names = tools.map((t) => t.name);
     const primary = names.filter((n) => n.startsWith("loopover_"));
     const legacy = names.filter((n) => n.startsWith("gittensory_"));
-    expect(primary.length).toBe(67);
+    expect(primary.length).toBe(68);
     expect(legacy.length).toBe(0);
-    expect(names.length).toBe(67);
+    expect(names.length).toBe(68);
   });
 
   it("no loopover_ tool's description carries a stale deprecation notice", async () => {
@@ -70,11 +71,11 @@ describe("MCP legacy alias retirement (#4777) — discovery invariants", () => {
     }
   });
 
-  it("`loopover-mcp tools --json` reports the same 67-tool count the live server registers", async () => {
+  it("`loopover-mcp tools --json` reports the same 68-tool count the live server registers", async () => {
     const { tools } = await client.listTools();
     const payload = JSON.parse(run(["tools", "--json"])) as { count: number; tools: Array<{ name: string }> };
     expect(payload.count).toBe(tools.length);
-    expect(payload.count).toBe(67);
+    expect(payload.count).toBe(68);
     expect([...payload.tools.map((t) => t.name)].sort()).toEqual([...tools.map((t) => t.name)].sort());
   });
 });
diff --git a/test/unit/routes-results-payload.test.ts b/test/unit/routes-results-payload.test.ts
new file mode 100644
index 0000000000..6723a49655
--- /dev/null
+++ b/test/unit/routes-results-payload.test.ts
@@ -0,0 +1,97 @@
+import { describe, expect, it } from "vitest";
+import { createApp } from "../../src/api/routes";
+import { buildResultsPayload, type IterationResult } from "../../src/results-payload";
+import { createTestEnv } from "../helpers/d1";
+
+// #6752: POST /v1/loop/results-payload — the REST mirror bringing loopover_build_results_payload to the same
+// parity its same-tier sibling loopover_check_slop_risk (/v1/lint/slop-risk) already has. The route delegates to
+// the pure buildResultsPayload (covered by its own unit tests in results-payload.test.ts), so these pin the ROUTE
+// contract: the composed payload is returned unmodified for every shape the MCP tool accepts, and a bad body is
+// rejected rather than passed through to the composer.
+const apiHeaders = (env: Env) => ({ authorization: `Bearer ${env.LOOPOVER_API_TOKEN}`, "content-type": "application/json" });
+const PATH = "/v1/loop/results-payload";
+
+const post = (env: Env, body: unknown) =>
+  createApp().request(PATH, { method: "POST", headers: apiHeaders(env), body: JSON.stringify(body) }, env);
+
+describe("POST /v1/loop/results-payload (#6752)", () => {
+  it("composes the customer-facing result for an iteration that opened a PR", async () => {
+    const env = createTestEnv();
+    const response = await post(env, {
+      repoFullName: "acme/widgets",
+      prNumber: 42,
+      title: "Add retry to the upload client",
+      changedFiles: [
+        { path: "src/upload.ts", additions: 12, deletions: 3 },
+        { path: "test/upload.test.ts", additions: 30, deletions: 0 },
+      ],
+      status: "merged",
+    });
+    expect(response.status).toBe(200);
+    await expect(response.json()).resolves.toMatchObject({
+      prLink: "https://github.com/acme/widgets/pull/42",
+      totals: { files: 2, additions: 42, deletions: 3 },
+    });
+  });
+
+  it("returns null prLink when the iteration opened no pull request", async () => {
+    const env = createTestEnv();
+    const response = await post(env, { repoFullName: "acme/widgets", prNumber: null, title: "Nothing to open" });
+    expect(response.status).toBe(200);
+    await expect(response.json()).resolves.toMatchObject({ prLink: null, totals: { files: 0, additions: 0, deletions: 0 } });
+  });
+
+  it("returns exactly what the pure composer returns for every shape the tool accepts", async () => {
+    const env = createTestEnv();
+    // One case per meaningful arm of the composer's input: PR/no-PR, absent vs null prNumber, each status,
+    // absent/empty/partial changedFiles, and a change over the diff-preview cap (totals must still count all).
+    const cases: IterationResult[] = [
+      { repoFullName: "acme/widgets", prNumber: 42, title: "Opened", status: "open" },
+      { repoFullName: "acme/widgets", prNumber: 42, title: "Merged", status: "merged" },
+      { repoFullName: "acme/widgets", prNumber: 42, title: "Closed", status: "closed" },
+      { repoFullName: "acme/widgets", prNumber: 7, title: "Status omitted defaults to open" },
+      { repoFullName: "acme/widgets", title: "prNumber absent entirely" },
+      { repoFullName: "acme/widgets", prNumber: null, title: "prNumber null" },
+      { repoFullName: "acme/widgets", prNumber: 9, title: "Empty changed set", changedFiles: [] },
+      { repoFullName: "acme/widgets", prNumber: 9, title: "Counts omitted", changedFiles: [{ path: "README.md" }] },
+      {
+        repoFullName: "acme/widgets",
+        prNumber: 9,
+        title: "Over the preview cap",
+        changedFiles: Array.from({ length: 12 }, (_, i) => ({ path: `src/f${i}.ts`, additions: i, deletions: 1 })),
+      },
+    ];
+    for (const body of cases) {
+      const response = await post(env, body);
+      expect(response.status, JSON.stringify(body)).toBe(200);
+      // PARITY: the route must return exactly what the pure composer the MCP tool calls returns.
+      await expect(response.json(), JSON.stringify(body)).resolves.toEqual(JSON.parse(JSON.stringify(buildResultsPayload(body))));
+    }
+  });
+
+  it("rejects an invalid or unparseable body with 400", async () => {
+    const env = createTestEnv();
+    for (const body of [
+      {},
+      { title: "missing repoFullName" },
+      { repoFullName: "", title: "empty repoFullName" },
+      { repoFullName: "acme/widgets" },
+      { repoFullName: "acme/widgets", title: 7 },
+      { repoFullName: "acme/widgets", title: "bad status", status: "reopened" },
+      { repoFullName: "acme/widgets", title: "bad prNumber", prNumber: 1.5 },
+      { repoFullName: "acme/widgets", title: "bad changedFiles", changedFiles: [{ additions: 1 }] },
+    ]) {
+      const response = await post(env, body);
+      expect(response.status, JSON.stringify(body)).toBe(400);
+      await expect(response.json()).resolves.toMatchObject({ error: "invalid_results_payload_request" });
+    }
+    const malformed = await createApp().request(PATH, { method: "POST", headers: apiHeaders(createTestEnv()), body: "{not json" }, createTestEnv());
+    expect(malformed.status).toBe(400);
+  });
+
+  it("leaks no wallet/hotkey/trust-score terms", async () => {
+    const env = createTestEnv();
+    const text = JSON.stringify(await (await post(env, { repoFullName: "acme/widgets", prNumber: 42, title: "Add retry" })).json());
+    expect(text).not.toMatch(/wallet|hotkey|coldkey|trust score|reward/i);
+  });
+});