Course 1.0: AI Captain for Claude Code, in progress Real code. Higher horizons.

Ship Better
Software, Faster.

AI Captain gives you specialist AI agents inside VS Code: code, test, debug, review, refactor and ship, all working together like a real crew.

See Course 1.0 Explore features
Runs on the tools you already have
VS Code GitHub Copilot Claude Code 1.0 MCP Tools MIT License
AI CAPTAIN
MBYou

/code add input validation to the signup form

AI Captain
  • Analyzing workspace… found 12 relevant files
  • Iteration 1: applied 3 changes in 2 files
  • Reviewing changes…
  • No critical issues found. Done!
2 files changed+60 -4
TSsignup.tssrc/auth+18 -4
TSsignup.test.tssrc/auth+42
Claude Opus
How it works

From Idea
to Launch

Your AI crew handles the heavy lifting, right inside VS Code.

  1. ☉ 1. Tell the Captain Describe what you want to build.
  2. ☸︎ 2. Your Crew Gets to Work Specialist agents plan, code, test and refine.
  3. ▣ 3. Review & Approve See changes, give feedback, iterate with ease.
  4. ⚓︎ 4. Ship It Production-ready code, faster than ever.
Features

Everything You Need to Ship Faster

Nine slash commands that supercharge your development workflow with AI-powered coding, testing, debugging, reviewing, and planning.

/c
Powered by GitHub Copilot Powered by Copilot

Autonomous Coding

Describe what you want, and AI Captain writes the code. Changes are automatically reviewed and refined until they pass quality checks.

@aicaptain /code
/r
AI Captain Expert Review

Code Review

Get instant, expert-level code reviews. Finds bugs, security issues (OWASP), performance problems, and suggests best practices.

@aicaptain /review
/t
AI Captain Test Writer

Test Generation

Detects your test framework, matches existing patterns, and generates unit, integration, and e2e tests that actually run.

@aicaptain /test
/d
AI Captain Debug Agent

Bug Tracing

Reads errors, traces call chains, identifies root causes, and proposes minimal fixes. Analysis-first approach.

@aicaptain /debug
How It Works

One Crew, One Workflow

AI Captain 0.5 orchestrates 19 specialist agents across your development workflow. Version 1.0 folds them into 8 sharper roles.

👨‍💻 You Describe the task
→
AI Captain Orchestrates
→
Copilot Generates code
Specialists Test, review, fix
→
🔄 Iterate Until perfect
→
✓ Done Code shipped

You Describe the Task

Type your request in Copilot Chat or the Activity Bar sidebar. AI Captain detects your intent automatically.

Copilot Generates Code

Using your workspace context, Copilot generates precise search-replace edits and applies them to your files.

Specialist Agents Take Over

Specialist agents handle testing, debugging, security auditing, code review, refactoring, and more, each optimized for its domain.

Iterate Until Perfect

If critical issues are found, AI Captain fixes them automatically. The loop continues until the code passes review.

@aicaptain /review

Code Review: src/auth.ts

⚠️ Security Issue: Password is not hashed before storage. Use bcrypt with at least 12 rounds.

🔎 Performance: Database query inside the loop causes N+1 problem. Batch with a single query.

✓ Best Practice: Good use of parameterized queries, preventing SQL injection.
AI CaptainExpert Review

Code Reviews That Find Real Issues

Unlike traditional linters, AI Captain's reviews check correctness, security (OWASP), types, architecture, performance, and style with a structured checklist.

  • ✓ OWASP Top 10 security vulnerability detection
  • ✓ Finds bugs, race conditions, and logic errors
  • ✓ Catches performance anti-patterns
  • ✓ Architecture and style recommendations
Course 1.0 In progress

A New Course: Claude Code

AI Captain is being rebuilt as one plugin that Claude Code and GitHub Copilot both load. A deterministic core does the checking, so the AI only handles what needs judgment.

  1. Code

    Facts

    Diff, typecheck, lint and test results, secrets and risky paths. The core gathers them before the model sees the task.

  2. AI

    Judgment

    The model writes and reviews the code and reports findings in a fixed schema: file, line, severity and evidence.

  3. Code

    Verdict

    Rules decide: GO or NO-GO before a commit, done or not done before the agent may stop. Same input, same verdict.

Done means the checks say so

When the agent wants to stop, a stop gate runs your project's typecheck and the related tests. Red sends it back to work with the exact file and line. Green lets it finish. After three attempts, or if the core itself fails, the gate steps aside and tells you.

That is the plan for 1.0, and the first spike already showed the gate holding in Claude Code.

claude · my-app
> add input validation to the signup form
… Claude edits signup.ts and wants to stop …
AI Captain: checks failed (attempt 1/3)
typecheck src/auth/signup.ts:42 TS2345
test     signup.test.ts "rejects empty email"
… Claude fixes both …
AI Captain: checks passed, gate open

The route

No release date yet. Each phase ends with a check that decides whether the next one starts.

  1. 0 Spike Hooks, skills, agents and MCP proven in Claude Code. The VS Code check comes next. Underway
  2. 1 Core and Claude Code Deterministic core, stop gate and pre-commit verdict. Planned
  3. 2 Crew and VS Code 19 agents become 8 roles. The same plugin runs in Copilot. Planned
  4. 3 Security, CI, workflows Scanners, a CI gate and verified multi-agent reviews. Planned
  5. 4 Companion and release A slim VS Code companion and version 1.0. Planned

The VS Code extension (0.5) is what the sections above describe. With 1.0 its logic moves into the plugin, and the extension becomes a slim companion with a status bar and a memory view.

Pricing

Free & MIT Licensed

AI Captain is free and MIT licensed. You bring the AI: GitHub Copilot today, Claude Code with 1.0.

+

Everything Included

Free

Everything in 0.5: 19 agents, 9 commands

  • ✓ /agent: Autonomous coding with tool access
  • ✓ /code: Coding with auto review loop
  • ✓ /test, /debug, /security: Specialist agents
  • ✓ /review, /research, /plan: Analysis tools
  • ✓ /precommit: Quality gate before commits
  • ✓ 19 installable specialist agents
  • ✓ Activity Bar sidebar
  • ✓ Persistent memory bank
  • ✓ MIT licensed
Follow Course 1.0

Not on the VS Code Marketplace yet.

FAQ

Frequently Asked Questions

Version 0.5 needs VS Code 1.99 or later and the GitHub Copilot Chat extension. Version 1.0 is planned as a plugin that runs in Claude Code as well as in VS Code with GitHub Copilot.

AI Captain is not on the VS Code Marketplace yet. The first public release is planned as version 1.0; the route above shows how it gets there.

Yes. AI Captain is free and MIT licensed. You bring the AI: a GitHub Copilot subscription today, Claude Code with 1.0.

Version 0.5 adds three things: an autonomous coding loop that applies and reviews changes, specialist agents for testing, debugging, security auditing and refactoring, and a memory bank that remembers decisions across sessions. Version 1.0 goes further: a deterministic core runs your project's checks and decides when work is done, instead of leaving that to the model.

AI Captain works with any language or framework your AI model handles: TypeScript, JavaScript, Python, Go, Rust, Java, C#, and many more. The 1.0 core is designed to find your project's typecheck, lint and test commands on its own.

Your code is processed by the AI model you already use: GitHub Copilot today, Claude Code with 1.0. AI Captain itself has no backend and no telemetry.

Version 0.5 includes agents for coding, testing, debugging, security auditing, code review, pre-commit review, refactoring, migration, docs writing, rapid MVP building, production readiness, multi-agent pipelines, full-stack development, architecture, performance, planning, TDD, build fixing and tech stack analysis. With 1.0 they become 8 roles with enforced tool rights: implementer, planner, reviewer, debugger, tester, security, refactorer and docs.

There is no date yet. Phase 0, a technical spike, is under way, and each phase only starts once the previous one has passed its checks.

Set Sail on the New Course

AI Captain 1.0 brings the crew to Claude Code and puts a deterministic core in charge of the checks.

See Course 1.0 Explore Features