Search Documentation

Find and navigate to documentation pages and their content

Quick Start Guide

Set up Codistry and begin coding with AI in minutes.

New to Codistry?

This guide assumes the extension is installed and your account is created. If not, start with the Installation Guide and then complete Account Setup.

With the free Developer tier active, connect an AI provider to begin.

Your AI Provider, Your Keys

The Developer tier uses your own API keys for OpenAI, Anthropic, or any OpenAI-compatible provider.

Your code and keys stay under your control. There are no relay servers and no extra charges beyond your provider's billing.

Everything is configured from the gear

Almost all of Codistry's configuration lives in the Settings view, open the Codistry sidebar and click the settings gear at the top. Models, providers, tool approvals, the Context Engine, and more are all there as simple toggles and inputs, so you rarely need to edit a config file by hand.

Step 1: Choose Your AI Provider

Open the Codistry sidebar in VS Code, then select the Settings gear at the top. You can connect to models in three ways:

OPTION AUse Anthropic Claude

Strong at code comprehension and careful instruction following.

  1. Get an API key from console.anthropic.com.
  2. In Settings → Models, enable Anthropic and paste your API key.
  3. In General settings, set Model Provider to Anthropic.
  4. Select a Claude Sonnet model for coding tasks.

OPTION BUse OpenAI

Access current GPT models such as GPT-5.5 and GPT-5.2, plus o-series reasoning models like o3.

  1. Get an API key from platform.openai.com.
  2. In Settings → Models, enable OpenAI and paste your API key.
  3. In General settings, set Model Provider to OpenAI.
  4. Choose a model from the dropdown, for example GPT-5.5, GPT-5.2, or a reasoning model such as o3.

OPTION CUse Local Models

Run models offline on your machine for maximum privacy.

  1. Install and run Ollama or LM Studio.
  2. In Settings → Models, enable Custom Models.
  3. Click + Add Model and configure:
    • Display Name: e.g. Local-OSS
    • URL: Ollama http://localhost:11434/v1, LM Studio http://localhost:1234/v1
    • Max Tokens: context window, e.g. 32000
    • Enable capabilities as supported (Tools, Streaming)
  4. In General settings, select your custom model from the dropdown.

For additional support with local model configuration, contact our team.

Want the full picture of built-in and custom models? See the Models guide →

Step 2: Your First Conversation

Start a Chat

  1. Open the Codistry sidebar.
  2. Type a message in the input box at the bottom.
  3. Press Enter or select the send button.

Try These Examples

Ask about a file:

What does src/index.ts do?

Request code changes:

Add error handling to the API routes

Search your codebase:

Find all files that import React

Get help with errors:

Why am I getting this TypeScript error?
Codistry can map your whole codebase

The first time you open a project, Codistry offers to index it so the agent understands how your code fits together, that's what powers questions like the ones above. Indexing runs locally and keeps itself up to date as you commit. The semantic layer downloads a small one-time local model and uses a few tokens to build.

It's on by default; you can build, re-run, or visualize indexes anytime in Settings → Context Engine. Learn more about the Context Engine →

Step 3: Understanding Tool Approvals

The agent asks for approval before it performs actions. You remain in control.

Tool Safety Levels

  • 🟢 Read Only: Safe to auto-approve. Examples: read_file, find_files_by_name, get_file_diagnostics.
  • 🟡 Modifying: Review carefully. Examples: edit_file, write_file, rename_symbol_across_codebase.
  • 🔴 Executing: Always review. Examples: run_shell_command, run_vscode_editor_command.

Tip: In Settings → Auto Approve Tools you can enable automatic approval for read-only tools.

Common Workflows

Understanding Code

Goal: Learn how a feature works.

Explain how the authentication system works.

What the agent does:

  1. Searches for authentication files.
  2. Reads relevant code.
  3. Analyzes the flow.
  4. Explains the implementation in plain language.

Debugging

Goal: Fix a failing test or error.

This test is failing: "should validate user input". Help me fix it.

What the agent does:

  1. Reads the test file.
  2. Checks diagnostics and logs.
  3. Identifies the likely cause.
  4. Proposes a fix and applies it with your approval.

Refactoring

Goal: Improve structure without changing behavior.

Refactor UserService to use dependency injection.

What the agent does:

  1. Reads the current implementation.
  2. Finds usages across the codebase.
  3. Proposes refactored code.
  4. Updates files systematically.

Documentation & Visualization

Goal: Generate documentation or diagrams.

Create a sequence diagram for the checkout flow.

What the agent does:

  1. Traces the code path.
  2. Identifies the key components.
  3. Generates a Mermaid diagram.
  4. Renders it in chat.

Explore More

Context Engine

Codebase-aware answers from a local map of your project.

Agents & Subagents

Delegate self-contained work to isolated subagents.

Personalization

Custom rules, adaptive learning, and project memory, all local.

Skills Runtime

Run task-specific code in a sandboxed Docker container.

MCP Extensions

Add custom tools using the Model Context Protocol.

Best Practices

Effective prompting and workflow optimization.

Setup complete

Codistry is ready to help. Ask questions about your code or request changes to start working faster.

Learn more: Best Practices • Settings Reference • FAQ

Previous← InstallationNextAccount & Subscription →