✨ Now available on VS Code Marketplace

AI coding agent with
visual clarity

ClarAIty brings a fully transparent AI coding agent directly into VS Code. Watch every tool call, trace every decision, and stay in control at every step.

↓ Install for VS Code View on GitHub
ClarAIty -- VS Code
Conversation
Refactor the auth module to use async/await
read_file src/auth.py
grep "def login"
edit_file src/auth.py
Done. Converted 3 functions to async. Tests still pass.
Agent running...
Live diff
- def login(user, password):
- result = db.query(user)
+ async def login(user, password):
+ result = await db.query(user)

+ if not result:
+ raise AuthError("Invalid credentials")
+ return create_session(result)
Tool trace
read_file -- src/auth.py (42 lines)
grep -- "def login" -- 3 matches
edit_file -- applied 7-line patch
run_command -- pytest tests/auth/ -- PASSED

Features
Everything you need, nothing hidden

ClarAIty is built on the principle that you should always know exactly what your AI agent is doing and why.

🔍

Full Tool Transparency

Every file read, every grep, every command -- displayed in real time as a live trace. No black box decisions.

🤖

Subagent Delegation

Delegate focused tasks to specialized subagents -- code reviewer, test writer, doc writer -- each with their own isolated context.

Approval Workflow

Destructive or sensitive operations pause for your approval. Auto-approve categories you trust, review the rest.

📄

Session History

Every conversation is persisted. Resume any past session, replay tool traces, and pick up exactly where you left off.

Undo Any Turn

Made a mistake? Undo the last agent turn -- file changes are snapshotted before every edit, so recovery is instant.

🔗

MCP Integration

Connect any Model Context Protocol server to extend the agent with custom tools, APIs, and data sources.

📚

Knowledge Base

Build a persistent knowledge graph of your codebase. Architecture, decisions, and invariants survive across sessions.

🌟

Task Tracker

Built-in beads task system. Create, prioritize, and track work items without leaving the agent chat.

🛠

Bring Your Own Model

Works with OpenAI, Anthropic, and any OpenAI-compatible endpoint. Configure model per session or per subagent.


How it works
From install to productive in minutes

No complex setup. Install the extension, add your API key, and start coding with AI that you can actually watch think.

1

Install the extension

One click from the VS Code Marketplace. The agent backend installs automatically via pip.

2

Add your API key

Paste your OpenAI or Anthropic key into settings. Keys are stored securely in VS Code SecretStorage -- never in files.

3

Open the sidebar

Click the ClarAIty icon in the activity bar. Type your first request and watch the agent get to work.

4

Stay in control

Review diffs before they land, approve sensitive operations, undo anything, and trace every decision.

Ready to code with clarity?

Free to install. Works with your existing API keys.

Install for VS Code -- Free View source on GitHub