# Advanced Usage

URL: https://docs.hanzo.ai/docs/dev/usage/advanced

> CI/CD integration, MCP configuration, verbose logging, and advanced workflows.

## Non-Interactive / CI Mode

Run Hanzo Dev headless in pipelines:

```yaml
- name: Update changelog via Hanzo Dev
  run: |
    npm install -g @hanzo/dev
    export OPENAI_API_KEY="${{ secrets.OPENAI_KEY }}"
    dev exec --full-auto "update CHANGELOG for next release"
```

### Resuming Non-Interactive Sessions

```bash
# Picker
dev resume

# Most recent
dev resume --last

# Specific session
dev resume <SESSION_ID>
```

## Tracing / Verbose Logging

Enable debug mode for detailed logging:

```bash
dev --debug "your prompt"
```

When `--debug` is enabled, request/response JSON is partitioned into helper-specific folders under `~/.hanzo/debug_logs/`:

- `auto/coordinator`
- `auto/observer/bootstrap`
- `auto/observer/cadence`
- `auto/observer/cross_check`
- `guided_terminal/agent_install_flow`
- `tui/rate_limit_refresh`
- `ui/theme_spinner`
- `cli/manual_prompt`

Monitor logs in real-time:

```bash
tail -F ~/.hanzo/debug_logs/codex-tui.log
```

Without `--debug`, only critical crash/error logs are written to `~/.hanzo/debug_logs/critical.log.*`.

## Model Context Protocol (MCP)

Configure MCP servers in `~/.code/config.toml`:

```toml
[mcp_servers.server-name]
command = "npx"
args = ["-y", "mcp-server"]
env = { "API_KEY" = "value" }
```

### Using Hanzo Dev as an MCP Server

Hanzo Dev can also run as an MCP server:

```bash
dev mcp
```

Test with the MCP inspector:

```bash
npx @modelcontextprotocol/inspector dev mcp
```

Send a `tools/list` request to see the available `dev` tool that accepts a catch-all of inputs including a `config` map for overrides.

### MCP Management

```bash
# Add a custom MCP server
dev mcp add my-search -- /path/to/search-mcp

# Add remote server with auth
dev mcp add api-tool --url https://api.example.com --bearer-token $TOKEN

# List configured servers
dev mcp list --json

# Remove a server
dev mcp remove my-search
```