Workflows

Workflows

Arxo MCP tools let an AI assistant run agent architecture analysis and work with arxo.yaml configuration. This page walks through practical workflows and shows the expected inputs and outputs.

Tip

Every tool that runs analysis accepts a project_path parameter. Use "." to analyze the project in the current workspace, or provide an absolute path.

---

Agent architecture analysis

When to use: To assess agent or LLM-powered code for reliability, governance, safety, and coordination. Use this when the codebase contains agents, tools, or LLM integrations.

Steps:

1. Run agent and OpenClaw architecture analysis:

analyze_architecture({
  "project_path": "."
})

2. The server returns metrics and findings for the agent_architecture metric. The response includes:

   • Composite scores (e.g. overall agent health, reliability, governance, safety, coordination)
   • Per-detector findings with evidence (file, line, reason)
   • UI schema with explanations and recommendations

3. The AI interprets the result and can suggest fixes (e.g. add loop guards, improve schema validation, add observability).

What to do next: Address high-severity findings first; use the recommendations in the response to prioritize changes.

---

Config help and validation

When to use: Before creating or editing an arxo.yaml file, or to validate an existing config.

Get config schema and examples

arxo_config_help({})

Or request a specific section:

arxo_config_help({ "section": "metrics" })

Sections: all, data, metrics, policy, report, examples.

Validate a config file

arxo_config_validate({ "project_path": "." })

Or validate inline YAML:

arxo_config_validate({
  "project_path": ".",
  "config_yaml": "metrics:\n  - id: agent_architecture\n    enabled: true"
})

The response lists errors and warnings (syntax, invalid metric IDs, etc.).

Suggest a config for the project

arxo_config_suggest({ "project_path": "." })

The server detects language and project layout and returns a suggested arxo.yaml you can paste into the project.

---

Performance

• Caching — Results from analyze_architecture are cached in memory with a configurable TTL (default 5 minutes). Repeated calls for the same project return cached results. Use --cache-ttl and --max-cache-entries in Configuration to tune.

Logging

The --log-level argument controls the MCP server log verbosity (e.g. --log-level debug). Logs are written to stderr. In Cursor, open the MCP server logs panel and select the Arxo server to inspect tool execution and any errors.