For AI agents: read the documentation index. Documentation pages are available as Markdown by appending .md to their URL, or requesting Accept: text/markdown.

Namera Docs

namera mcp serve

Start the local stdio MCP server for an agent client.

Start the local stdio MCP server for an agent client.

Usage

namera mcp serve --profile default

Output

This is a long-running protocol process, not a command that prints a result. Stdout contains MCP messages only; diagnostics and authorization instructions go to stderr. The client starts and stops the process. Closing stdin ends it.

Options

--profile

  • Type: string
  • Default: default

Choose the MCP connection name. This is separate from ordinary CLI profiles, whose default is personal. Use the same name for serve, login, status, and logout.

--host

  • Type: absolute API URL
  • Default: https://api.namera.ai

Choose the API origin for this MCP connection. The origin and profile together identify its saved credentials. Use HTTPS, or a loopback HTTP origin for local development. NAMERA_API_URL does not override this option on MCP commands.

--max-gas-cost-wei

  • Type: non-negative integer string, in wei
  • Default: sponsored operations only

Approve a maximum gas cost for self-funded executions. MCP tools cannot raise this process-level ceiling. Omit it unless you intend to allow the agent to pay gas from the account.

namera mcp serve --profile default --max-gas-cost-wei 1000000000000000

--output and --quiet do not turn the stdio protocol into ordinary command output. Do not redirect stdout into a log file.

Authorization

Register this command with your MCP client instead of running a daemon manually. The first authenticated tool call starts browser authorization when needed. After approving it, retry the tool. Ordinary CLI login and NAMERA_API_KEY do not authorize MCP.

Supported protocol versions are 2026-07-28, 2025-11-25, and 2025-06-18. Negotiation is automatic; there is no protocol-version flag.

Errors

Failures exit nonzero and are written to stderr. With --output json, inspect error.code and error.nextStep. See Output and Errors.