Skip to main content
Glama
progress_visibility.md3.53 kB
# Real-time Tool Progress Visibility (EX MCP Server) This document explains how EX MCP Server emits real-time progress breadcrumbs during tool execution and how you can see them in VS Code (Augment Code) and Auggie CLI. ## What changed - Added a centralized progress helper `utils/progress.py` with: - `send_progress(message: str, level: str = "info")` for emitting progress - `set_mcp_notifier(...)` hook for future MCP-native notifications - Default streaming mode controlled by `STREAM_PROGRESS` (default: true) - Wrapper now defaults `LOG_LEVEL=INFO` when `AUGGIE_CLI=true` or `STREAM_PROGRESS=true` - Workflow tools (analyze, thinkdeep, codereview, debug, precommit, refactor, secaudit, planner, consensus, testgen, tracer) emit progress at key milestones via BaseWorkflowMixin - Simple tools (like chat) also emit minimal progress via SimpleTool ## How it works - Tools call `send_progress("message")` at key milestones - Messages go to the `mcp_activity` logger at INFO level as `[PROGRESS] ...` - MCP notifier hook (best-effort) is registered on server startup to allow future MCP-native notifications ## How to see progress ### VS Code (Augment Code) - Open the Output panel or the extension’s tool activity channel while a tool runs - Ensure the server is started with: - `STREAM_PROGRESS=true` (default) - Optionally `AUGGIE_CLI=true` (enables INFO by default in the wrapper) - Look for lines like: - `mcp_activity - INFO - [PROGRESS] analyze: Starting step 1/3 - ...` ### Auggie CLI - Auggie launches the server via our wrapper; with `AUGGIE_CLI=true`, `LOG_LEVEL=INFO` is set by default - Watch the CLI output or tail logs/logs/mcp_server.log for `[PROGRESS]` lines ### Logs - Always available at `logs/mcp_server.log` - Windows PowerShell: `Get-Content logs\mcp_server.log -Wait -Tail 50` ## Direct tool testing (no MCP client) Use the included tool bench or in-process invocation to see progress: - Planner (no model required): ```bash python scripts/tool_bench.py run planner --json "{\"step\":\"Create a small plan\",\"step_number\":1,\"total_steps\":1,\"next_step_required\":false,\"findings\":\"Testing\"}" ``` - ThinkDeep/Analyze (require models): set provider keys first, then run similar commands. Quick smoke test runner: ```bash python scripts/progress_test.py ``` This calls planner, thinkdeep, and analyze in-process via `server.handle_call_tool(...)` and prints results while the server logs `[PROGRESS]` breadcrumbs. ## Environment variables - `STREAM_PROGRESS` (default: `true`): master switch for emitting progress - `AUGGIE_CLI` (default: unset): when `true`, wrapper sets `LOG_LEVEL=INFO` - `LOG_LEVEL` (default: `INFO` server-side): controls verbosity - `LOG_FORMAT` (default: `json` server-side): set to `plain` for human-readable output ## Current limitations and next steps - MCP-native notifications: the helper exposes `set_mcp_notifier`; when the python MCP SDK provides a stable notification API (e.g., `window/logMessage` or `$\progress`), we’ll wire it for richer UI surfacing in clients. - Model-required tools won’t emit progress if they fail model validation early. Configure provider keys or use `planner` to validate progress in isolation. ## FAQ - Q: I don’t see `[PROGRESS]` lines in VS Code. - Ensure `STREAM_PROGRESS=true` and wrapper default to `INFO` is in effect or set `LOG_LEVEL=INFO` explicitly. - Check the extension Output channel and logs/mcp_server.log. - Q: Can I disable progress? - Set `STREAM_PROGRESS=false`.

Latest Blog Posts

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/Zazzles2908/EX_AI-mcp-server'

If you have feedback or need assistance with the MCP directory API, please join our Discord server