feat: mcp server for code interpreter

integrations/code-interpreter-mcp/Dockerfile

@@ -0,0 +1,25 @@
+# Copyright The Volcano Authors.
+#
+# SPDX-License-Identifier: Apache-2.0
+
+# Build from repository root:
+#   docker build -f integrations/code-interpreter-mcp/Dockerfile -t agentcube-code-interpreter-mcp:latest .
+
+FROM python:3.12-slim
+
+WORKDIR /app
+
+COPY sdk-python /build/sdk-python
+COPY integrations/code-interpreter-mcp /build/mcp
+
+RUN pip install --no-cache-dir /build/sdk-python \
+    && pip install --no-cache-dir /build/mcp \
+    && rm -rf /build
+
+ENV MCP_TRANSPORT=streamable-http \
+    MCP_HOST=0.0.0.0 \
+    MCP_PORT=8000
+
+EXPOSE 8000
+
+ENTRYPOINT ["python", "-m", "agentcube_code_interpreter_mcp", "--transport", "streamable-http", "--host", "0.0.0.0", "--port", "8000"]

integrations/code-interpreter-mcp/README.md

@@ -0,0 +1,141 @@
+# AgentCube Code Interpreter MCP Server
+
+[MCP](https://modelcontextprotocol.io/) server that exposes the AgentCube code interpreter (Router + Workload Manager) as tools for Cursor, Claude Desktop, and other hosts.
+
+**You need a running AgentCube cluster and a `CodeInterpreter` CR before this server is useful.** Follow the steps below in order.
+
+---
+
+## 0. Deploy AgentCube (same path as the project tutorial)
+
+Do this once on a Kubernetes cluster (v1.24+). Full detail: [`docs/getting-started.md`](../../docs/getting-started.md).
+
+1. **agent-sandbox** — install CRDs + controller (see getting-started **Step 1**).
+2. **Redis** — create namespace `agentcube`, deploy Redis ( **Step 2** ).
+3. **AgentCube** — `git clone` the repo, then `helm install agentcube ./manifests/charts/base ...` ( **Step 3** ).
+4. **CodeInterpreter** — e.g. `kubectl apply -f example/code-interpreter/code-interpreter.yaml` ( **Step 4** ). That manifest creates `my-interpreter` in namespace **`default`**.
+
+Check:
+
+```bash
+kubectl get pods -n agentcube
+kubectl get codeinterpreter
+```
+
+---
+
+## 1. Install the MCP server (your machine)
+
+Python 3.10+. The server calls the Python SDK at runtime, so install **both** the SDK and this package.
+
+**From a clone of this repository** (recommended while developing):
+
+```bash
+cd /path/to/agentcube
+pip install -e ./sdk-python
+pip install -e ./integrations/code-interpreter-mcp
+```
+
+**Released SDK only** (if you publish or install `agentcube-sdk` from PyPI): install the SDK, then install this package from the same repo revision so versions stay compatible.
+
+---
+
+## 2. Point the server at Router + Workload Manager
+
+From your laptop, port-forward (same pattern as [getting-started **Step 5**](../../docs/getting-started.md)):
+
+```bash
+# Terminal A
+kubectl port-forward -n agentcube svc/workloadmanager 8080:8080
+
+# Terminal B — local 8081 → Router service port 8080 (avoids clashing with 8080)
+kubectl port-forward -n agentcube svc/agentcube-router 8081:8080
+```
+
+Then:
+
+```bash
+export WORKLOAD_MANAGER_URL="http://localhost:8080"
+export ROUTER_URL="http://localhost:8081"
+```
+
+If your cluster enforces auth for those APIs, set `API_TOKEN` (Bearer) the same way as the Python SDK.
+
+---
+
+## 3. Run the MCP server
+
+**stdio** (typical for Cursor / Claude Desktop):
+
+```bash
+python -m agentcube_code_interpreter_mcp --transport stdio
+```
+
+**Streamable HTTP** (debugging; default path `/mcp` per the Python MCP SDK):
+
+```bash
+python -m agentcube_code_interpreter_mcp --transport streamable-http --host 0.0.0.0 --port 8000
+```
+
+---
+
+## 4. Example: Cursor `mcp.json`
+
+Match **namespace** and **interpreter name** to your `CodeInterpreter` CR. For `example/code-interpreter/code-interpreter.yaml` (`my-interpreter` in `default`):
+
+```json
+{
+  "mcpServers": {
+    "agentcube-code-interpreter": {
+      "command": "python",
+      "args": ["-m", "agentcube_code_interpreter_mcp", "--transport", "stdio"],
+      "env": {
+        "ROUTER_URL": "http://localhost:8081",
+        "WORKLOAD_MANAGER_URL": "http://localhost:8080",
+        "AGENTCUBE_NAMESPACE": "default",
+        "CODE_INTERPRETER_NAME": "my-interpreter"
+      }
+    }
+  }
+}
+```
+
+If the CR name is not `my-interpreter`, set `CODE_INTERPRETER_NAME` accordingly. Keep the two `kubectl port-forward` terminals open while you use the IDE.
+
+---
+
+## 5. Optional: run the MCP server inside the cluster
+
+Build from **repository root** (Docker context must include `sdk-python`):
+
+```bash
+docker build -f integrations/code-interpreter-mcp/Dockerfile -t agentcube-code-interpreter-mcp:latest .
+kubectl apply -f integrations/code-interpreter-mcp/deployment.yaml
+```
+
+Edit `deployment.yaml`: `ROUTER_URL` / `WORKLOAD_MANAGER_URL` must match your Services (defaults target `agentcube` namespace). Set `CODE_INTERPRETER_NAME` to your CR name.
+
+**Replicas:** with `session_reuse=true`, clients are cached in process memory — use **one** replica or sticky routing at your gateway.
+
+---
+
+## Configuration reference
+
+| Variable | Description |
+|----------|-------------|
+| `ROUTER_URL` | **Required.** Router base URL |
+| `WORKLOAD_MANAGER_URL` | **Required.** Workload Manager base URL |
+| `API_TOKEN` | Optional Bearer token (same as SDK) |
+| `AGENTCUBE_NAMESPACE` | Namespace for API paths; default `default` |
+| `CODE_INTERPRETER_NAME` | CR name; default `my-interpreter` |
+| `CODE_INTERPRETER_SESSION_TTL` | New session TTL (seconds); default `3600` |
+| `MCP_TRANSPORT` | `stdio` or `streamable-http` (or `--transport` on CLI) |
+| `MCP_HOST` / `MCP_PORT` | Listen address for HTTP mode |
+
+---
+
+## Tools
+
+Names match **`CodeInterpreterClient`**: `run_code`, `execute_command`, `write_file`, `list_files`, `upload_file`, `download_file`, and `stop_session` (call when finished after `session_reuse=true`). Optional `session_id` / `session_reuse` on tools.
+
+`upload_file` / `download_file` paths are on the **host that runs** the MCP process (not inside the sandbox).

integrations/code-interpreter-mcp/agentcube_code_interpreter_mcp/__init__.py

@@ -0,0 +1,10 @@
+# Copyright The Volcano Authors.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+
+"""AgentCube Code Interpreter MCP server."""
+
+from agentcube_code_interpreter_mcp.server import create_mcp_server
+
+__all__ = ["create_mcp_server"]

integrations/code-interpreter-mcp/agentcube_code_interpreter_mcp/__main__.py

@@ -0,0 +1,41 @@
+# Copyright The Volcano Authors.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+
+"""CLI entry for AgentCube Code Interpreter MCP server."""
+
+from __future__ import annotations
+
+import argparse
+import os
+
+from agentcube_code_interpreter_mcp.server import create_mcp_server
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(description="AgentCube Code Interpreter MCP server")
+    parser.add_argument(
+        "--transport",
+        choices=["stdio", "streamable-http"],
+        default=os.environ.get("MCP_TRANSPORT", "stdio"),
+        help="MCP transport (default: stdio, or MCP_TRANSPORT)",
+    )
+    parser.add_argument(
+        "--host",
+        default=os.environ.get("MCP_HOST", "127.0.0.1"),
+        help="Listen host for streamable-http (default: MCP_HOST or 127.0.0.1)",
+    )
+    parser.add_argument(
+        "--port",
+        type=int,
+        default=int(os.environ.get("MCP_PORT", "8000")),
+        help="Listen port for streamable-http (default: MCP_PORT or 8000)",
+    )
+    args = parser.parse_args()
+    mcp = create_mcp_server(host=args.host, port=args.port)
+    mcp.run(transport=args.transport)
+
+
+if __name__ == "__main__":
+    main()