Skip to content
| Marketplace
Sign in
Visual Studio Code>AI>AISHA DirigentNew to Visual Studio Code? Get it now.
AISHA Dirigent

AISHA Dirigent

Evymo s.r.o.

|
39 installs
| (0) | Free
AI-powered development flow moderator — orchestrates code quality, compliance, testing, and task routing via MCP Knowledge Server and n8n workflows. Works with any project.
Installation
Launch VS Code Quick Open (Ctrl+P), paste the following command, and press enter.
Copied to clipboard
More Info

AISHA Dirigent — Autonomous Development Orchestrator for VS Code

Proactive AI development orchestrator that watches your workspace and autonomously guides development decisions, code quality, compliance, and testing. Connects to MCP Knowledge Server and n8n agent workflows.

Two modes in one extension: autonomous Dirigent Brief (triggered on file save) and manual @aisha chat (on demand).

Works with any project — connect to your own backend or use the hosted AISHA Cloud platform.

Interaction Modes

Autonomous Mode — Dirigent Brief System

When you save a file, the extension analyzes your workspace state and sends a structured Dirigent Brief to the default Copilot agent (regular chat, not @aisha). This enables the agent to actually fix code, not just advise.

The brief includes:

  • Errors — active TypeScript/ESLint diagnostics for that file
  • Development State — file role, missing tests, migration status
  • Workflow Directives — prioritized next steps (test-first: missing test is always step 1)
  • KB Rules — relevant architecture rules from the knowledge base for that file role
  • Story Context — active project, ruleset, and compliance requirements

File roles detected: hook, component, schema, migration, i18n, test, page

On first brief per session you get a notification with three options:

  • Send to Chat — open brief in Copilot chat once
  • Always send — auto-approve for the rest of the session
  • Dismiss — skip this brief

Manual Mode — @aisha Chat Participant

Ask questions, request code reviews, or trigger agent workflows on demand via VS Code Chat:

@aisha /test src/hooks/useMyHook.ts
@aisha /compliance check this migration
@aisha /onboard
@aisha what story am I working on?

Slash commands:

Command Description
/test Generate or review tests for the active file
/quality Code quality and architecture review
/compliance Check against KB rules and security patterns
/estimate Effort estimate for a task or story
/next Suggest next development step
/health Platform health and agent status
/route Route task to a specific model (n8n, MCP fallback)
/models Browse AI model registry and usage stats
/eval Evaluate agent output or story delivery
/proposals List open proposals for the active story
/instructions Sync copilot-instructions.md from KB
/dirigent Direct Dirigent orchestration prompt
/onboard Full autonomous project onboarding pipeline
/story Query or manage the active story context
/local Chat using a local LLM (edge-first mode)
/connect-repo Connect workspace to a GitHub repository

Getting Started

  1. Install the extension from a .vsix file
  2. Run AISHA Dirigent: Connect (Setup Wizard) from the Command Palette
    • Step 1: Select backend profile (local / cloud / custom)
    • Step 2: Authenticate (skipped for local dev)
    • Step 3: Select or enter your active story
    • Step 4: Auto-sync copilot-instructions.md from KB
  3. Save any file to receive your first Dirigent Brief
  4. Or start manually with @aisha in VS Code Chat

Configuration

All settings are under aisha.dirigent.* (VS Code settings or .aisha/dirigent.json):

Setting Description
aisha.dirigent.mcpUrl MCP Knowledge Server URL
aisha.dirigent.supabaseUrl AISHA API / Gateway URL (PostgREST style) for auth, push channel, context sync and RPC calls. Uses the current stack (local warmup or self-hosted), not legacy Supabase.
aisha.dirigent.anonKey Public anon key for the gateway (from local-warmup or prod .env)
aisha.dirigent.n8nTriggerUrl n8n trigger URL for agent orchestration
aisha.dirigent.storyId Default story UUID (overridden by .aisha/story.json)
aisha.dirigent.expertiseLevel Guidance profile — controls depth of decision support: beginner (Educating) / intermediate (Collaborative) / advanced (Autonomous) / expert (Supervisory)

MCP, n8n, RPC, push, and context sync calls use the signed-in AISHA ID / Keycloak access token. Static MCP or service-role tokens are not supported by the extension.

Config resolution order (highest priority first):

  1. AISHA_* environment variables
  2. .aisha/dirigent.local.json (gitignored, per-machine overrides)
  3. .aisha/dirigent.json (committed workspace config)
  4. VS Code settings fallback

Connection Profiles

Named backend profiles in .aisha/dirigent.json let you switch environments:

{
  "profiles": {
    "local": { "supabaseUrl": "http://127.0.0.1:8000", "mcpUrl": "..." },   // from `npm run warmup:local` (gateway)
    "selfhosted": { "supabaseUrl": "https://api.yourdomain.com", "mcpUrl": "..." }
  },
  "activeProfile": "local"
}

Switch with AISHA Dirigent: Switch Backend or edit activeProfile directly.

Commands

Command Description
AISHA Dirigent: Connect (Setup Wizard) 4-step wizard: profile → auth → story → sync
AISHA Dirigent: Login Authenticate with your account
AISHA Dirigent: Logout Sign out
AISHA Dirigent: Create Account (Signup) Register a new account
AISHA Dirigent: Switch Backend Toggle between configured backend profiles
AISHA Dirigent: Set Active Story Set project context
AISHA Dirigent: Set Guidance Profile Adjust guidance depth and decision support level
AISHA Dirigent: Sync Copilot Instructions Generate copilot-instructions.md from KB
AISHA Dirigent: Refresh AI Models Reload model registry
AISHA Dirigent: Clear Active Session Reset session state

Architecture

VS Code workspace
├── Dirigent Brief (auto-flow.ts)          ← file save triggers state assessment
│   ├── detectFileRole()                   ← hook / schema / migration / component…
│   ├── assessDevState()                   ← errors, missing tests, story context
│   ├── buildDirigentBrief()               ← structured prompt with directives + KB rules
│   └── sendBriefToChat()                  ← to DEFAULT Copilot agent (not @aisha)
│
├── @aisha Chat Participant (participant.ts) ← manual mode, slash commands
│   └── routes to MCP + n8n workflows
│
├── AISHA Push ←── SSE ←── backend        ← real-time story/decision events
├── Session Manager                        ← story context, decision log, run_id
└── Status Bar                             ← backend, session, expertise
        │
        ├── MCP Knowledge Server (JSON-RPC 2.0 / svc-mcp-knowledge)
        │       └── search_knowledge, get_expert_rule, validate_compliance…
        │
        └── n8n Orchestration
                └── Knowledge, Compliance, Delivery, Dirigent, Ragnarok agents

Key distinction: Dirigent Briefs go to regular Copilot chat (agent can edit code directly). Manual @aisha requests go through the chat participant (advisory, tool calls, structured responses).

Development

cd extensions/aisha-dirigent
npm install
npm run compile
# Press F5 in VS Code to launch Extension Development Host

Build .vsix

npx @vscode/vsce package --no-dependencies

License

Source code is licensed under the Elastic License 2.0 (ELv2). Copyright © 2026 Evymo s.r.o. and AISHA Platform contributors.

You may use, copy, modify, and deploy this extension (including for your clients). You may not offer it to third parties as a hosted or managed service, remove the licensing/copyright notices, or circumvent any license-key functionality.

The hosted AISHA service (internal models, evaluation and routing) is a separate proprietary service operated by Evymo s.r.o. and is not part of, nor licensed under, ELv2. "AISHA", "Evymo" and "Dirigent" are trade names of Evymo s.r.o.; no trademark rights are granted by the source license.

  • Contact us
  • Jobs
  • Privacy
  • Manage cookies
  • Terms of use
  • Trademarks
  • Your Privacy Choices
  • Consumer Health Privacy
© 2026 Microsoft