diff --git a/README.md b/README.md
index b4ba74c..1f05f4c 100644
--- a/README.md
+++ b/README.md
@@ -241,6 +241,37 @@ mneme is a standard **MCP server**, supporting both **stdio** (default, one proc
 | Windsurf | Add to MCP server config |
 | Cline / Continue | Add to MCP settings |
 
+**Config provided, not verified by us:**
+
+| Agent | Setup |
+|-------|-------|
+| DeepSeek Harness (dsh) | Copy [`examples/dsh.cordis.yml`](examples/dsh.cordis.yml) into your dsh config — see below |
+
+### DeepSeek Harness
+
+dsh deliberately ships no memory layer of its own. Its design notes reject
+per-provider plugins outright, on the grounds that they repeat "auth,
+configuration, lifecycle, and tool wrappers that MCP already standardizes" — so
+the supported path is a Cordis overlay pointing its generic MCP client at
+whatever server you want.
+
+That means mneme needs no dsh plugin and no adapter code. It already speaks
+stdio MCP, which is the only thing the overlay requires. The integration is a
+config file: [`examples/dsh.cordis.yml`](examples/dsh.cordis.yml).
+
+Two caveats worth reading before you wire it up:
+
+- **The overlay uses `node <abs-path>/mcp-server.mjs`, not a bare command.**
+  mneme is not on npm with a `bin`, so there is nothing on PATH to invoke. You
+  need a checkout.
+- **Set `EMBEDDING_API_*`.** Without them mneme starts, answers, and gives you
+  FTS5 keyword search only. Hybrid recall is the reason to run it; losing it is
+  not an error, just a quieter and worse result set.
+
+The overlay is provided as a starting point and is **not covered by this repo's
+test suite** — the CI here exercises mneme's own MCP server, not dsh's loader.
+Reports of it working (or not) are welcome.
+
 ---
 
 ## Features
diff --git a/examples/dsh.cordis.yml b/examples/dsh.cordis.yml
new file mode 100644
index 0000000..6ed9303
--- /dev/null
+++ b/examples/dsh.cordis.yml
@@ -0,0 +1,41 @@
+# DeepSeek Harness (dsh) overlay for mneme — opt-in, default-off.
+#
+# mneme needs no dsh plugin: it is already a standard MCP server and dsh ships a
+# generic MCP client (@deepseek-ai/dsh-mcp-client). This file only points that
+# client at a mneme checkout. Copy it into your dsh config and edit the path.
+#
+# Unlike the overlays dsh ships in examples/mcp-memory, `command` here is `node`
+# with an absolute script path rather than a bare executable: mneme is not
+# published to npm with a `bin`, so there is nothing on PATH to invoke. Clone the
+# repo, run `npm install`, and point args at that checkout.
+#
+# Two things that will silently give you a worse system if you skip them:
+#
+#   1. TOKENMEM_DB_PATH. Without it the DB is created relative to the process
+#      working directory, so the store you build in one session can be invisible
+#      from the next. Set it absolute.
+#
+#   2. EMBEDDING_API_*. Without these mneme starts fine and answers queries — as
+#      FTS5 keyword search only. Hybrid recall (FTS + vector KNN fused by RRF) is
+#      the reason to run mneme at all, and its absence is not an error, just a
+#      quieter, worse result set. Recent builds warn once per process on the
+#      degrade path; check stderr on first run if you are unsure which mode you
+#      are in.
+- insert:
+    - id: memory-mneme
+      name: '@deepseek-ai/dsh-mcp-client'
+      config:
+        serverName: mneme
+        transport: stdio
+        command: node
+        # EDIT THIS PATH to your mneme checkout.
+        args: ['/absolute/path/to/mneme/mcp-server.mjs']
+        cwd: !!js process.cwd()
+        env:
+          TOKENMEM_DB_PATH: !!js >-
+            process.env.TOKENMEM_DB_PATH?.trim() || process.getBuiltinModule('node:path').join(process.getBuiltinModule('node:os').homedir(), '.mneme', 'mneme.db')
+          # Optional but strongly recommended — see note 2 above.
+          EMBEDDING_API_BASE_URL: !!js process.env.EMBEDDING_API_BASE_URL ?? ''
+          EMBEDDING_API_KEY: !!js process.env.EMBEDDING_API_KEY ?? ''
+          EMBEDDING_MODEL: !!js process.env.EMBEDDING_MODEL ?? 'text-embedding-3-small'
+          EMBEDDING_DIMENSION: !!js process.env.EMBEDDING_DIMENSION ?? '1536'