Chat Participant

The @aicoach chat participant gives you conversational access to all AI Engineer Coach data directly in the VS Code chat panel. Ask questions in natural language and get data-driven coaching responses without leaving your editor.

Getting Started

Before using @aicoach, open the AI Engineer Coach sidebar at least once to load your session data.

Note: When you use @aicoach, your question and the results returned by AI Engineer Coach tools are sent to the selected VS Code chat model so it can synthesize a response. The underlying coaching data is gathered locally, but the final chat answer is not produced purely through local processing.

Type @aicoach in any VS Code chat panel followed by your question:

@aicoach how am I doing this week?

The participant is sticky — once invoked, follow-up messages in the same thread continue the conversation without needing to type @aicoach again.

Slash Commands

Command	Description	Default prompt
/summary	Quick usage overview	Highlights strengths and top areas to improve
/improve	Improvement recommendations	Top 3 things to improve with specific actions
/compare	Tool comparison	Compare AI coding tools and their effectiveness
/flow	Flow & focus analysis	Deep work patterns and best productivity hours

Use a slash command with no additional text to get the default analysis, or add your own question:

@aicoach /flow Am I more productive in the morning or afternoon?

Available Tools

The participant has access to 12 backend tools that it selects automatically based on your question:

Tool	Domain	What it returns
aiEngineerCoach_summary	Observe	Session counts, recommendations, top anti-patterns
aiEngineerCoach_activity	Observe	Daily requests, LOC, sessions, and harness breakdown
aiEngineerCoach_credits	Measure	AI credit usage, per-model breakdown, daily trend, and costly requests
aiEngineerCoach_codeProduction	Measure	AI vs user LOC, language breakdown, workspace distribution
aiEngineerCoach_flow	Measure	Deep work scores, best hours, follow-up latency
aiEngineerCoach_patterns	Improve	Anti-patterns and practice recommendations with severity
aiEngineerCoach_insights	Improve	Learning velocity, intent classification, prompt maturity
aiEngineerCoach_wellbeing	Improve	Work-life balance score, time distribution, burnout risk
aiEngineerCoach_workflows	Improve	Repeated workflow clusters with automation suggestions
aiEngineerCoach_harnessComparison	Observe	Side-by-side tool comparison: sessions, LOC, cancel rates
aiEngineerCoach_sessions	Observe	Browse or search individual sessions by ID or keyword
aiEngineerCoach_contextHealth	Improve	Context utilization, compaction, config health, and instruction quality

All tools accept optional fromDate, toDate, workspaceId, and harness filters. The participant resolves relative time references ("last week", "past month") automatically.

How It Works

The participant runs an agentic loop that:

1. Sends your question along with a coaching persona and tool-selection heuristics to the language model
2. The model decides which tools to call based on your intent
3. Tool results are fed back into the conversation for the model to synthesize
4. The model may call additional tools if needed (up to 8 rounds)
5. A final, synthesized coaching response is streamed back to you

This means a single question like "compare my productivity this week vs last week" can trigger multiple tool calls (activity, flow, code production) and produce a unified answer.

Example Conversations

Broad check-in:

@aicoach Give me a quick health check

→ Calls summary, returns practice scores, session count, top anti-pattern, and a suggested next step.

Specific investigation:

@aicoach Why is my prompt quality score dropping?

→ Calls patterns with recent date range, surfaces the specific anti-patterns driving the score down with example prompts from your sessions.

Follow-ups

After each response, the participant suggests follow-up prompts to guide deeper analysis:

• Improve — "What should I improve next?"
• Compare tools — "Compare my AI tools"
• Flow state — "How is my focus & flow?"

Click any follow-up to continue the conversation without typing.