Documentation Index
Fetch the complete documentation index at: https://docs.meshai.dev/llms.txt
Use this file to discover all available pages before exploring further.
MeshAI MCP Server
The MeshAI MCP Server brings governance and observability to any MCP-compatible AI tool — even tools that don’t support custom API base URLs.
When to Use
| Integration Path | Best For |
|---|
| MeshAI Proxy | Tools with custom base URL support (Claude Code, Cursor, Cline, Codex) |
| MeshAI MCP Server | Tools without proxy support (Claude Desktop, VS Code, JetBrains, Windsurf, Zed, Warp) |
| Both | Maximum coverage — proxy for traffic monitoring, MCP for governance tools |
Setup
No global install needed — each tool config uses npx to fetch and run the server directly from GitHub.
Environment Variables
| Variable | Required | Default | Description |
|---|
MESHAI_API_KEY | Yes | — | Your MeshAI API key (msh_...) |
MESHAI_API_URL | No | https://api.meshai.dev | API endpoint |
MESHAI_AGENT_NAME | No | mcp-agent | Name for this agent in the registry |
Claude Desktop
Edit ~/Library/Application Support/Claude/claude_desktop_config.json (Mac) or %APPDATA%\Claude\claude_desktop_config.json (Windows):
{
"mcpServers": {
"meshai": {
"command": "npx",
"args": ["-y", "@meshailabs/mcp-server"],
"env": {
"MESHAI_API_KEY": "msh_your_key_here",
"MESHAI_AGENT_NAME": "claude-desktop"
}
}
}
}
Cursor
Add to .cursor/mcp.json or Settings > MCP Servers:
{
"meshai": {
"command": "npx",
"args": ["-y", "@meshailabs/mcp-server"],
"env": {
"MESHAI_API_KEY": "msh_your_key_here",
"MESHAI_AGENT_NAME": "cursor"
}
}
}
Windsurf
Edit ~/.codeium/windsurf/mcp_config.json:
{
"mcpServers": {
"meshai": {
"command": "npx",
"args": ["-y", "@meshailabs/mcp-server"],
"env": {
"MESHAI_API_KEY": "msh_your_key_here",
"MESHAI_AGENT_NAME": "windsurf"
}
}
}
}
VS Code (GitHub Copilot)
Add to .vscode/mcp.json:
{
"servers": {
"meshai": {
"command": "npx",
"args": ["-y", "@meshailabs/mcp-server"],
"env": {
"MESHAI_API_KEY": "msh_your_key_here",
"MESHAI_AGENT_NAME": "vscode"
}
}
}
}
JetBrains IDEs
Settings > Tools > AI Assistant > Model Context Protocol (MCP) > Add Server:
- Command:
npx
- Args:
-y @meshailabs/mcp-server
- Environment:
MESHAI_API_KEY=msh_your_key_here, MESHAI_AGENT_NAME=jetbrains
Works with IntelliJ IDEA, PyCharm, WebStorm, GoLand, Rider, and all other JetBrains IDEs.
Zed
Add to ~/.config/zed/settings.json:
{
"context_servers": {
"meshai": {
"source": "custom",
"command": {
"path": "npx",
"args": ["-y", "@meshailabs/mcp-server"]
},
"env": {
"MESHAI_API_KEY": "msh_your_key_here",
"MESHAI_AGENT_NAME": "zed"
}
}
}
}
Warp
Settings > Agent > MCP > Add Server:
- Command:
npx
- Args:
-y @meshailabs/mcp-server
- Environment:
MESHAI_API_KEY=msh_your_key_here, MESHAI_AGENT_NAME=warp
AWS Kiro
Add to .kiro/mcp.json:
{
"mcpServers": {
"meshai": {
"command": "npx",
"args": ["-y", "@meshailabs/mcp-server"],
"env": {
"MESHAI_API_KEY": "msh_your_key_here",
"MESHAI_AGENT_NAME": "kiro"
}
}
}
}
Cline / Roo Code / Kilo Code
Settings > MCP Servers > Add Server:
- Command:
npx
- Args:
-y @meshailabs/mcp-server
- Environment:
MESHAI_API_KEY=msh_your_key_here, MESHAI_AGENT_NAME=cline
Continue.dev
Add to ~/.continue/config.yaml:
mcpServers:
- name: meshai
command: npx
args:
- "-y"
- "@meshailabs/mcp-server"
env:
MESHAI_API_KEY: msh_your_key_here
MESHAI_AGENT_NAME: continue
Amazon Q Developer
Add to ~/.aws/amazonq/mcp.json:
{
"mcpServers": {
"meshai": {
"command": "npx",
"args": ["-y", "@meshailabs/mcp-server"],
"env": {
"MESHAI_API_KEY": "msh_your_key_here",
"MESHAI_AGENT_NAME": "amazon-q"
}
}
}
}
Gemini CLI
Add to ~/.gemini/settings.json:
{
"mcpServers": {
"meshai": {
"command": "npx",
"args": ["-y", "@meshailabs/mcp-server"],
"env": {
"MESHAI_API_KEY": "msh_your_key_here",
"MESHAI_AGENT_NAME": "gemini-cli"
}
}
}
}
Sourcegraph Cody / Tabnine / Goose
These tools follow similar MCP configuration patterns. Set:
- Command:
npx
- Args:
-y @meshailabs/mcp-server
- Environment:
MESHAI_API_KEY=msh_your_key_here, MESHAI_AGENT_NAME=<tool-name>
Refer to each tool’s MCP documentation for the exact config file location.
Once configured, the MCP server exposes 8 tools to the AI assistant:
| Tool | Description |
|---|
meshai_register | Register this tool as a governed agent in MeshAI |
meshai_heartbeat | Send health heartbeat for uptime monitoring |
meshai_track_usage | Report token usage for cost tracking |
meshai_check_policy | Check if a model/provider is allowed before use |
meshai_get_risk | Get EU AI Act risk classification |
meshai_get_compliance | Get compliance readiness score (0-120) |
meshai_report_incident | Report a security incident (Article 73) |
meshai_audit_log | Query the governance audit trail |
Available Resources
| URI | Description |
|---|
meshai://agent/status | Current agent status and metadata |
meshai://governance/policies | Active governance policies |
meshai://compliance/score | EU AI Act readiness score |
How It Works
- Your AI tool (e.g., Claude Desktop) starts the MeshAI MCP server via
npx
npx fetches the package from GitHub, builds it, and runs it as a subprocess- The server connects to the MeshAI API using your API key
- The AI tool can call MeshAI governance tools alongside its normal capabilities
- The MeshAI dashboard shows this tool alongside all other agents
- Governance policies, audit trails, and compliance features apply universally
| Tool | MCP Support | Proxy Support | MCP Server Fills Gap? |
|---|
| Claude Desktop | Yes | No | Yes |
| Claude Code | Yes | Yes | Optional |
| Cursor | Yes | Yes | Optional |
| Windsurf | Yes | Partial | Yes |
| Cline | Yes | Yes | Optional |
| Roo Code | Yes | Yes | Optional |
| Kilo Code | Yes | Yes | Optional |
| AWS Kiro | Yes | TBD | Yes |
| VS Code Copilot | Yes | No | Yes |
| JetBrains IDEs | Yes | No | Yes |
| Zed | Yes | No | Yes |
| Warp | Yes | No | Yes |
| Continue.dev | Yes | Yes | Optional |
| Sourcegraph Cody | Yes | Yes | Optional |
| Tabnine | Yes | No | Yes |
| Amazon Q Developer | Yes | No | Yes |
| Gemini CLI | Yes | No | Yes |
| Goose | Yes | No | Yes |
| Codex | Yes | Yes | Optional |
“MCP Server Fills Gap?” — Yes means the tool has no custom base URL support, so the MCP server is the only way to get MeshAI governance. Optional means the tool already supports the proxy; the MCP server adds governance tools but isn’t required.
