Agents
Use Mezcal with coding agents through the CLI and hosted MCP transport.
Loading documentation content…
Use Mezcal with coding agents through the CLI and hosted MCP transport.
Use the Agents lane when a human or model is operating Mezcal through tools instead of raw REST requests.
Mezcal currently exposes two agent-facing paths:
For most agent users, the shortest path is:
mezcal/api basemezcal mcp serve --transport remoteShared environment:
export MEZCAL_BASE_URL="https://<your-mezcal-host>/api"
export MEZCAL_API_KEY="mzk_test_your_key_here"
export MEZCAL_CHAIN="SN_MAIN"If your agent client wants a JSON config instead of a one-shot command, this is the minimum useful shape:
{
"mcpServers": {
"mezcal": {
"command": "mezcal",
"args": ["mcp", "serve", "--transport", "remote"],
"env": {
"MEZCAL_BASE_URL": "https://<your-mezcal-host>/api",
"MEZCAL_API_KEY": "${MEZCAL_API_KEY}",
"MEZCAL_CHAIN": "SN_MAIN"
}
}
}
}That is the practical parent-page answer for Claude Code, Codex-adjacent setups, Cursor-like MCP clients, and other tool-calling environments.
When you are replacing lightweight chain reads with an agent, start with:
statustoken_total_supplytoken_balance_oftoken_transfersThose give the agent a narrow, typed Mezcal surface before it starts guessing at raw chain structure.
They solve different problems:
That separation is intentional. It keeps the local developer flow fast while giving remote agent consumers a safer, more explicit auth and policy boundary.