Lobstack is a managed cloud platform for deploying autonomous AI agents. Each agent runs on a dedicated virtual machine with persistent memory, 700+ integrations, and goes live in under 90 seconds with zero DevOps.

How long does it take to deploy an AI agent?

You can deploy an AI agent on Lobstack in under 90 seconds. Describe what your agent should do, pick a model, and Lobstack handles all the infrastructure — no Docker, Kubernetes, or server setup needed.

What AI models does Lobstack support?

Lobstack supports 16 AI models including Claude Opus and Sonnet from Anthropic, GPT-5 and GPT-4o from OpenAI, Gemini 2.5 Pro from Google, plus DeepSeek, Mistral, and Grok. You can hot-swap models without losing agent memory or configuration.

Is Lobstack secure for enterprise use?

Yes. Every agent runs in a gVisor-sandboxed VM with AES-256 encryption at rest, TLS 1.3 in transit, HashiCorp Vault for secrets, row-level security on all tables, and full audit logging. Lobstack is SOC 2 aligned.

Docs/OpenClaw/MCP Servers

Guide

MCP Servers

Model Context Protocol (MCP) servers extend your agent with standardized tools from a 13,000+ server ecosystem. Connect to filesystems, APIs, databases, and more.

What is MCP?#

The Model Context Protocol (MCP) is an open standard originally developed by Anthropic for connecting AI agents to external tools and data sources. It was donated to the Linux Foundation's Agentic AI Foundation in late 2025 and is co-governed by Anthropic, OpenAI, and Block.

MCP servers are lightweight programs that expose tools (functions the AI can invoke), resources (data for context), and prompts (templates). They communicate via JSON-RPC 2.0 over stdio or HTTP.

Portable

MCP servers work with any MCP-compatible host: OpenClaw, Claude Code, Cursor, VS Code, and more.

Standardized

JSON-RPC 2.0 protocol with typed schemas. No custom integration code needed.

Ecosystem

13,000+ community servers on registries like npm, PyPI, and mcp.run.

Available MCP Servers#

Lobstack pre-configures these MCP servers. Enable them from Dashboard → Skills → MCP.

Server	Package	Category	Description
Filesystem	@modelcontextprotocol/server-filesystem	Filesystem	Read, write, search files in the agent workspace
Web Fetch	@modelcontextprotocol/server-fetch	Search	Fetch and extract content from URLs
GitHub	@modelcontextprotocol/server-github	Development	Interact with repos, issues, PRs, and actions
Memory	@modelcontextprotocol/server-memory	Data	Persistent key-value memory across conversations
Slack	@modelcontextprotocol/server-slack	Communication	Read and send Slack messages, manage channels

How It Works#

When an MCP server is enabled, OpenClaw spawns it as a child process and discovers its available tools. The AI model can then call those tools naturally during conversation, just like built-in tools.

MCP Server Lifecycle

1. Agent starts  ->  OpenClaw reads openclaw.json
2. For each mcpServers entry:
   a. Spawn process (npx/node/python command)
   b. Initialize JSON-RPC 2.0 connection
   c. Discover available tools via "tools/list"
   d. Register tools in agent's tool inventory
3. During conversation:
   a. AI model decides to call an MCP tool
   b. OpenClaw forwards the call to the server
   c. Server executes and returns results
   d. Results passed back to the AI model

Configuration#

MCP servers are configured in the mcpServers section of openclaw.json:

openclaw.json — MCP section

{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "/root/.openclaw/workspace"]
    },
    "github": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-github"],
      "env": {
        "GITHUB_PERSONAL_ACCESS_TOKEN": "ghp_your_token_here"
      }
    }
  }
}

Configuration Fields#

Field	Type	Description
command	string	Command to spawn the server (npx, node, python, uvx)
args	string[]	Arguments passed to the command
env	object	Environment variables for the server process

MCP vs Skills#

Aspect	MCP Servers	OpenClaw Skills
Protocol	JSON-RPC 2.0 (standardized)	Markdown + YAML frontmatter
Runtime	External process (child process)	Runs inside OpenClaw gateway
Portability	Works with any MCP host	OpenClaw-specific
Ecosystem	13,000+ servers	5,700+ on ClawHub
Use Case	Structured tool access	Natural language task guidance
Best For	APIs, databases, file systems	Workflows, specialized tasks

✅

Use both

MCP servers and Skills complement each other. Use MCP for structured tool access (GitHub API, database queries) and Skills for natural-language task guidance (code review workflows, content writing patterns).

Security#

MCP servers run on the agent VM with these security measures:

Process isolation — Each MCP server runs as a separate child process
Workspace scoping — Filesystem server restricted to /root/.openclaw/workspace
Credential management — API keys stored via Lobstack's encrypted credential system
Audit logging — MCP server enable/disable actions are logged

⚠️

Community servers

Before enabling a community MCP server, review its source code and check for known vulnerabilities. Only install servers from trusted publishers.

API Reference#

List MCP Servers#

GET /api/agent/mcp

curl -b "sb-access-token=YOUR_TOKEN" \
  "https://lobstack.ai/api/agent/mcp"

Enable/Disable MCP Server#

POST /api/agent/mcp

curl -X POST -H "Content-Type: application/json" \
  -b "sb-access-token=YOUR_TOKEN" \
  -d '{"serverId": "github", "enabled": true, "env": {"GITHUB_PERSONAL_ACCESS_TOKEN": "ghp_..."}}' \
  "https://lobstack.ai/api/agent/mcp"

OpenClaw Integration Channels Overview