A3Code λ — AI Coding Agent

Welcome to A3Code λ, your intelligent, spec-driven AI coding partner designed to plan, build, reason, and execute across modern software workflows — from dashboards and research papers to cloud, Kubernetes, and GitHub automation.

🚀 Get Started in 60 Seconds

Before anything else, do this 👇

1. Click the ⚙️ Settings cog
2. Enter your LLM Provider API key
3. Choose one of the supported providers (both offer free API keys):

• 👉 https://tensorix.ai
• 👉 https://openrouter.ai

Once your key is set, you’re ready to code with A3Code λ.

🧠 How to Use A3Code λ Effectively

1️⃣ Plan → Then Build

A3Code λ works best when you separate thinking from execution:

• Plan Mode — design, reason, and decompose work
• Build Mode — write code, run commands, and modify files

Always good to start in Plan Mode, then switch to Build Mode. You will find the switch is defaulted to build mode in the 👉 text prompt input. Click on the build switch and change it to 1️⃣ plan 1️⃣ mode.

2️⃣ Spec-Driven Development (Recommended)

A3Code λ supports spec-driven development.

• Write or reuse a spec
• Run it directly from:
  👉 https://a3code.ai

The agent will interpret the spec, generate code, and refine until the spec is satisfied.

3️⃣ Building a Dashboard?

Just ask the agent to:

Show it

The running application will be rendered in the internal browser.

📊 Financial & Research Power Features

🔌 APEXE3 MCP Server

Enable in Settings to unlock:

• Financial market datasets
• Research-oriented workflows
• Quantitative analysis pipelines

📈 FDS Mode (Financial Data Scientist)

Enable FDS Mode in Settings to analyze research papers and generate code.

Example:

/paper https://arxiv.org/pdf/2505.05325

A3Code λ will analyze the paper and produce runnable, high-quality code.

⌨️ Key Prompt Commands

You can control sessions and context directly from the prompt:

• /new — start a brand new session
• /sessions — list previous sessions and resume one
• /reduce — summarize progress and reduce the context window
• /context <directory> — add a directory to the agent’s working context

These commands help keep long-running projects fast and focused.

💡 Pro Tips for Power Users

• 📁 One project = one directory = one VS Code
• 🧩 Use a separate VS Code window per project
• ⏸️ If the agent thinks too long, stop it and say “continue”
• 🔄 Ask it to run servers (Node.js, APIs, dashboards) in the background
• 🔪 It can:
  • Kill processes
  • Check which process is listening on a port
• 🧑‍💻 GitHub automation:
  • Commits & PRs supported if git / gh CLI is installed
• ☸️ Kubernetes management:
  • Works with kubectl
• ☁️ Cloud operations:
  • Works with installed CLIs like gcloud, aws, az

⚙️ A3Code Agent Runtime (Important)

• A3Code agentic logic runs as a background process
• Default port: 9876
• All VS Code windows share this single process
• The process starts and stops automatically when the extension is:
  • Enabled
  • Disabled
  • Installed
  • Uninstalled

🛑 If the process hangs

On rare occasions, the process may not shut down cleanly.

Run:

ps -ef | grep ae3code

Then identify the PID and run:

kill -9

Restart VS Code with the A3Code AI Agent enabled.

You can also find ae3code in:

• Windows → Task Manager
• macOS → Activity Monitor

✨ Why A3Code λ?

• Plan-first, execution-second intelligence
• Spec-driven by design
• Research → Code workflows
• Financial data + AI engineering in one agent
• Deep local system integration

🧭 What to Do Next

1. Open Settings
2. Add your API key
3. Start in Plan Mode
4. Build something ambitious 🚀

A3Code λ doesn’t just write code — it understands your intent.

Happy building.