New Features in v0.1.31

This document outlines the major features added in indusagi-coding-agent v0.1.31.

Quick Start

• MCP Guide - Learn how to use Model Context Protocol for external integrations
• Memory Guide - Learn how Memory helps the assistant remember context

[PKG] Feature Summary

1. Model Context Protocol (MCP)

Connect your AI assistant to external tools and services.

What it enables:

• Access external APIs and tools
• Filesystem operations
• [GitHub] GitHub integration
• [Search] Web search and scraping
• [DB] Database queries
• [Web] Real-time web access

Quick setup:

# Create ~/.indusagi/mcp-servers.json
indusagi
# Tools are automatically available!

Examples:

User: "Read my README.md"
# Uses filesystem MCP

User: "Search GitHub for open issues"
# Uses GitHub MCP

User: "What's the latest Node.js update?"
# Uses web search MCP

Read Full MCP Guide →

---

2. Memory System

The assistant now remembers important context across sessions.

What it enables:

• [Storage] Persistent conversation history
• 🧠 Semantic understanding of past decisions
• Context-aware responses
• Faster problem-solving
• Project knowledge accumulation

Automatic setup: Memory works out of the box. Just start using indusagi!

Examples:

Session 1:
User: "I prefer TypeScript for all projects"

Session 2 (days later):
User: "Create a new API"
Assistant: "I'll use TypeScript as you prefer..."
# Assistant remembers your preference!

Read Full Memory Guide →

---

Feature Comparison

Feature	Before v0.1.31	After v0.1.31
External Tools	Limited to built-in tools	Full MCP ecosystem
GitHub Integration	Manual instructions	Automatic via MCP
Web Access	Web fetch only	Web search, fetch, scraping
Context Memory	Session-only	Persistent across sessions
Semantic Search	Not available	Full vector-based search
Database Access	No direct access	Multiple DB systems supported

---

Configuration Files

MCP Configuration

Location: ~/.indusagi/mcp-servers.json

{
  "servers": {
    "filesystem": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "$HOME"]
    }
  }
}

Memory Configuration

Location: ~/.indusagi/memory.json

{
  "enabled": true,
  "storage": "in-memory",
  "vectorStore": "in-memory",
  "maxMemoryItems": 1000,
  "similarityThreshold": 0.7
}

---

🛠️ Common Use Cases

Use Case 1: Code Review with Context

User: "Review my TypeScript code. Remember, we prefer functional patterns."

Memory: Recalls your functional programming preference
MCP: Accesses your GitHub to fetch the code
Result: Context-aware, personalized review

Use Case 2: Project Management

User: "Update the project status"

Memory: Recalls previous project decisions
MCP: Accesses GitHub to update issues
Result: Consistent with past architectural decisions

Use Case 3: Research with Memory

Session 1: "We're building a real-time chat app"
Session 2: "What database should we use?"

Memory: Recalls the chat app context
MCP: Searches for latest solutions
Result: Recommendations tailored to your use case

---

Architecture

MCP Integration

indusagi
  ↓
MCP Client Pool
  ├─→ Filesystem Server
  ├─→ GitHub Server
  ├─→ Web Search Server
  └─→ Custom Servers

Memory Integration

Conversation
  ↓
Memory System
  ├─→ Store as vector embedding
  ├─→ Search for relevant context
  └─→ Inject into conversation
  ↓
Enhanced Response

---

[Config] Environment Variables

MCP

INDUSAGI_DEBUG=1              # Enable MCP debug output
MCP_CONFIG_PATH=~/.indusagi/  # Custom config location

Memory

MEMORY_ENABLED=true           # Enable/disable memory
MEMORY_THRESHOLD=0.7          # Similarity threshold
OPENAI_API_KEY=sk-...         # For semantic embeddings

---

Integration with Other Features

GPT-5.4 Models

v0.1.31 includes full support for GPT-5.4:

• gpt-5.4 - Base model
• gpt-5.4-codex - Coding optimized (default for openai-codex)

These work seamlessly with MCP and Memory!

Clean CLI Output

MCP debug messages are suppressed by default for a clean terminal experience.

Enable with: INDUSAGI_DEBUG=1 indusagi

---

Getting Started

Step 1: Update to v0.1.31

npm install -g indusagi-coding-agent@0.1.31
indusagi --version
# Should show: 0.1.31

Step 2: Enable MCP (Optional)

# Copy example config
cp ~/.npm-global/lib/node_modules/indusagi-coding-agent/examples/mcp-servers.example.json ~/.indusagi/mcp-servers.json

# Edit to customize
nano ~/.indusagi/mcp-servers.json

Step 3: Set Environment Variables

export GITHUB_TOKEN="your_token"
export BRAVE_API_KEY="your_key"
export OPENAI_API_KEY="sk-..."

Step 4: Start Using!

indusagi

Memory and MCP are ready to use!

---

Documentation Index

Topic	Document
MCP Setup & Usage	MCP.md
Memory System	MEMORY.md
Installation	../README.md
Troubleshooting	../README.md#troubleshooting

---

Tips & Tricks

MCP Tips

• Start with filesystem and web search for maximum value
• Set up GitHub if you work with repos
• Use environment variables for secrets

Memory Tips

• Be explicit about preferences and constraints
• Reference past decisions to strengthen memory
• Start sessions with relevant context
• Search memory with indusagi --search-memory "query"

---

[Q&A] FAQ

Q: Is my memory data stored online?
A: No! Memory is stored locally in ~/.indusagi/memory/. Only OpenAI embeddings API may be called (if enabled).

Q: Can I use MCP without Memory?
A: Yes! They work independently. Enable only what you need.

Q: Does Memory slow down indusagi?
A: Minimal impact (<5% overhead). Memory retrieval is <100ms.

Q: Can I share Memory across machines?
A: Export and import: indusagi --export-memory and indusagi --import-memory file.json

Q: How often should I clear Memory?
A: You don't need to! Memory automatically manages old items. Clear only if privacy is a concern.

---

Getting Help

If you encounter issues:

1. MCP Issues: See MCP.md Troubleshooting
2. Memory Issues: See MEMORY.md Troubleshooting
3. General Issues: Check ../README.md

---

Version: 0.1.31
Released: March 2026
Status: · Production Ready