Frequently Asked Questions

1. Open VS Code
2. Extensions (Ctrl/Cmd + Shift + X) → search Codistry
3. Click Install

See the Installation Guide. For offline setups, see Air-Gapped Installation.

1. Create an account and sign in (Account Setup)
2. Enter your AI provider key(s) in Settings
3. Open Codistry from the sidebar and start a chat

Walkthrough: Quick Start.

Yes. The Developer tier is free. You bring your own AI keys and pay your provider directly. Full access to tools and features.

• Generate code: “Create a React UserCard component.”
• Explain code: “Walk me through parseConfig().”
• Debug:“Help fix this TypeError.”
• Refactor: “Convert callbacks to async/await in AuthService.”
• Tests:“Add unit tests for priceCalculator.ts.”
• With tools: Use MCP integrations naturally.

• Active file is visible automatically
• Selection → select code, then ask
• Copy/paste into chat
• @-mention files to reference them

Yes. Changes are previewed for review before applying, and you can undo any edit.

Tools power capabilities like reading files, searching code, or running commands. The agent chooses tools based on your request.

Recommendation: Auto-approve read-only tools. Keep writes and run_shell_command manual. Configure in Settings → Approval Settings.

Yes. Install via VSIX and configure local or on-prem model endpoints. See Air-Gapped Installation.

Yes. Use custom OpenAI-compatible endpoints or local runtimes such as Ollama and LM Studio. Add them via the adronite-agent.customModels array in Settings → Custom Models.

Your code is never sent to Codistry servers, only to your configured AI provider. In air-gapped deployments, nothing leaves your network.

• Anthropic: Claude Opus 4.8, Claude Sonnet 4.6, Claude Haiku 4.5
• OpenAI: GPT-5.5, GPT-5.2, o3, o4-mini
• Local: Via Ollama, LM Studio, or other OpenAI-compatible servers
• Custom: Any OpenAI-compatible API endpoint

See the in-extension model dropdown for the current list of available models.

• Claude Sonnet 4.6: Best balance of speed and quality
• Claude Opus 4.8: Deep analysis + long contexts
• GPT-5.5 / GPT-5.2: Fast, accurate generalist
• o3 / o4-mini: Advanced reasoning tasks
• Local models: Maximum privacy; offline friendly

Use the model picker in the chat header. That choice applies to the current session only, and each chat keeps its own model.

The model set in Settings → Models is the default that every new chat starts with. Changing it doesn't switch chats you've already started. See Models → Default model.

The agent learns your global preferences (tone, code style, workflow) across projects. Enable in Settings → Context. Learnings are stored locally.

Project Memory (Beta) builds a persistent knowledge graph for your workspace: entities, relationships, and patterns that survive across chats. Unlike Adaptive Learning (global), it's per-workspace.

Skills execute specialized tasks inside isolated Docker containers (e.g., document processing, data analysis). Requires Docker.

See the Skills Runtime Guide.

MCP (Model Context Protocol) adds custom tools (databases, APIs, scripts) that the AI can call. Available to all users.

Learn more about MCP Extensions →

Open Settings → Usage for local analytics: work acceleration, activity stats, and charts over 24h / 7d / 30d / 90d.

All metrics are computed and stored locally.

Settings → Account to view license status, plan info, and renewal date. Click Manage Subscription for the billing portal.

1. Check internet (if not air-gapped)
2. Verify API key(s) in Settings
3. Open Output/Logs for errors
4. Reload VS Code window
5. Update to the latest version

• Invalid key: Re-enter the key and restart
• Rate limit/quota: Check your provider dashboard
• Wrong model name: Use the exact identifier
• Network policies: Corporate proxies/firewalls may block requests