The sandbase-harness Model Context Protocol (MCP) server provides seamless integration between AI assistants (such as Claude Desktop, Cursor, and Claude Code) and Open-source CMA-compatible agent runtime for any model, with MCP tools, sandboxed sessions, audit, replay, and a local console. Includes a native DeepSeek Harness bundle over stdio MCP. By connecting this server, your AI agent gains specialized capabilities to inspect, query, and automate tasks within the AI/LLM domain.
Once installed in Claude Desktop or Cursor, try asking your assistant:
Copy the config for your editor. Verify environment variables or setup notes in the README.
Add to claude_desktop_config.json:
{
"mcpServers": {
"sandbase-harness": {
"command": "npx",
"args": [
"-y",
"sandbaseai/sandbase-harness"
]
}
}
}
A local-first runtime for AI agents. Sessions, sandboxed tools, memory, credentials, audit trails, and a built-in Console — all running on your machine or in your own infrastructure. `bash git clone --branch v0.3.1 --depth 1 https://github.com/sandbaseai/sandbase-harness.git cd sandbase-harness npm ci