Add MCP server — expose brain (recall/remember/ingest/stats) to any MCP host

Makes SecondBrain a first-class connector for Claude Desktop, Open WebUI,
LibreChat, Cursor, and any MCP-speaking client. Optional requirements-mcp.txt.
Atlas Corporation proprietary.

README.md

@@ -138,6 +138,27 @@ docker run --rm -v sbdata:/data -e OLLAMA_URL=http://host.docker.internal:11434
   secondbrain recall "quarterly numbers" --hybrid
 ```
 
+## Plug into any AI (MCP server)
+
+Expose your brain as tools to any MCP host — **Claude Desktop, Open WebUI, LibreChat, Cursor**,
+etc. — so the assistant gains local, private, long-term memory:
+
+```bash
+pip install "mcp[cli]"
+python3 mcp_server.py           # stdio MCP server
+```
+
+Add to your MCP host's config:
+```json
+{ "mcpServers": { "secondbrain": {
+    "command": "python3", "args": ["/opt/secondbrain/mcp_server.py"],
+    "env": { "BRAIN_HOME": "/opt/secondbrain/data", "OLLAMA_URL": "http://localhost:11434" } } } }
+```
+
+Tools exposed: **`recall`** (hybrid search of your memory), **`remember`** (store a note, scrubbed),
+**`ingest_path`** (index a folder), **`stats`**. This is how SecondBrain becomes a first-class
+*connector* in a Claude-style product surface — the same way MCP servers extend Claude.
+
 ## Use it as a library
 
 ```python

mcp_server.py

@@ -0,0 +1,90 @@
+#!/usr/bin/env python3
+# SecondBrain — (c) 2026 Atlas Corporation. All Rights Reserved. Proprietary; see LICENSE.
+"""SecondBrain MCP server — expose your brain as tools to any MCP host.
+
+Add this server to Claude Desktop, Open WebUI, LibreChat, Cursor, or any MCP client,
+and the assistant gains long-term, local, privacy-scrubbed memory + recall.
+
+Run (stdio transport, the default MCP wiring):
+    pip install "mcp[cli]"
+    python3 mcp_server.py
+
+Example Claude Desktop / MCP host config:
+    {
+      "mcpServers": {
+        "secondbrain": {
+          "command": "python3",
+          "args": ["/opt/secondbrain/mcp_server.py"],
+          "env": { "BRAIN_HOME": "/opt/secondbrain/data",
+                   "OLLAMA_URL": "http://localhost:11434" }
+        }
+      }
+    }
+
+Tools exposed: recall · remember · ingest_path · stats.
+Config is the same env as the engine (BRAIN_HOME, OLLAMA_URL, …).
+"""
+import os, sys
+sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
+import brain
+
+try:
+    from mcp.server.fastmcp import FastMCP
+except Exception:
+    sys.stderr.write("SecondBrain MCP needs the MCP SDK:  pip install \"mcp[cli]\"\n")
+    raise
+
+mcp = FastMCP("secondbrain")
+
+
+@mcp.tool()
+def recall(query: str, k: int = 6, hybrid: bool = True) -> list:
+    """Search long-term memory (the Second Brain). Returns the top-k most relevant
+    stored snippets with their source, project, timestamp and relevance score.
+    Use this to remember prior decisions, facts, documents, emails, code, and context.
+    `hybrid=True` fuses semantic + exact-keyword search (best for names/IDs/paths)."""
+    fn = brain.brain_recall_hybrid if hybrid else brain.brain_recall
+    return fn(query, k)
+
+
+@mcp.tool()
+def remember(text: str, project: str = "note", ref: str = "manual") -> str:
+    """Store a note/fact in long-term memory. Credentials are scrubbed automatically.
+    It becomes searchable immediately (lexical) and semantically after the next embed."""
+    conn = brain.db()
+    n = brain._stage(conn, "note", project, "", ref, "", "note", text)
+    conn.commit(); conn.close()
+    try:
+        brain.embed(limit=64)
+    except Exception:
+        pass
+    return f"stored {n} new chunk(s)"
+
+
+@mcp.tool()
+def ingest_path(path: str) -> str:
+    """Index a local directory of files (code/docs/notes) into memory, then embed."""
+    p = os.path.expanduser(path)
+    if not os.path.isdir(p):
+        return f"not a directory: {p}"
+    brain.ingest_files(p)
+    try:
+        brain.embed(limit=20000)
+    except Exception:
+        pass
+    return f"ingested + embedded from {p}"
+
+
+@mcp.tool()
+def stats() -> str:
+    """Report memory coverage: total chunks, embedded count, and sources."""
+    conn = brain.db()
+    tot = conn.execute("SELECT count(*) FROM chunks").fetchone()[0]
+    emb = conn.execute("SELECT count(*) FROM chunks WHERE embedded=1").fetchone()[0]
+    bysrc = dict(conn.execute("SELECT source,count(*) FROM chunks GROUP BY source").fetchall())
+    conn.close()
+    return f"chunks={tot} embedded={emb} by_source={bysrc}"
+
+
+if __name__ == "__main__":
+    mcp.run()

requirements-mcp.txt

@@ -0,0 +1,3 @@
+# Optional — only needed to run the MCP server (mcp_server.py).
+# The engine, wizard and connectors do not require this.
+mcp[cli]>=1.2