# Davmunrey Trazum MCP server

Source: https://glimind.com/server/mcp-registry%2Fio.github.davmunrey%2Ftrazum
Registry id: `mcp-registry/io.github.davmunrey/trazum`
Status: unknown

## Install

```json
{
  "mcpServers": {
    "davmunrey-trazum-mcp-server": {
      "command": "npx",
      "args": [
        "-y",
        "@trazum/mcp"
      ]
    }
  }
}
```

Claude Code:

```sh
claude mcp add davmunrey-trazum-mcp-server -- npx -y @trazum/mcp
```

## Tools (12)

- `analyze_task`: Analyze explicit native task transcripts for recorded usage and evidence.
- `check_prompt`: Check if a prompt fits maxTokens and if optimizing would help.
- `compare_tasks`: Compare two saved task reports using criteria, outcomes and evidence gates.
- `import_usage`: Convert native transcript text into normalized JSONL usage log.
- `list_models`: Return prices, context windows, cacheable minimums and review date.
- `optimize_prompt`: Return shorter text, token counts, monthly cost difference, and advisories.
- `plan_savings`: Rank savings actions from usage log, keeping assumptions and evidence separate.
- `position`: Show month's spend against ceilings from log text alone.
- `profile_usage`: Show spend breakdown from usage log text: per label, model, caching value.
- `prompt_writer`: Interview user with fixed questions and write a prompt from their answers.
- `spend_guard`: Check if a planned call fits the budget; return yes/no/cannot-tell.
- `verify_savings`: Check saved plan JSON against newer log for arrived/not-arrived actions.

Measured by Glimind (https://glimind.com): MCP handshake + tool list only, no tool is ever called. CC BY 4.0.