Agent-Fusion
🥈SilverAgent Fusion is a local RAG semantic search engine that gives AI agents instant access to your code, documentation (Markdown, Word, PDF). Query your codebase from code agents without hallucinations. Runs 100% locally, includes a lightweight embedding model, and optional multi-agent task orchestration. Deploy with a single JAR
6530Updated 2mo ago
Intermediate30min to implementautomation
Saves ~240 min per use
Quick InstallView Source
git clone https://github.com/krokozyab/Agent-Fusion.gitWorks with:
Claude
Overview
About This Skill
Agent Fusion is a local RAG semantic search engine designed to give AI coding agents instant, accurate access to your codebase and documentation without hallucinations. It indexes code files, PDFs, Word documents, and Markdown with embeddings, then enables AI assistants to query semantically—finding "authentication logic" even when implemented as "login", "credentials", or "access control". The system runs 100% locally as a single JAR with a lightweight embedded model, no external APIs or cloud dependencies required. An optional Task Manager component enables multi-agent collaboration with voting and decision tracking. Agents query via the query_context tool, receiving results ranked across semantic, symbol-based, and full-text search methods, with hierarchical relationships for traversing code structures and call chains.
How to Use
1. **Deploy Agent-Fusion**: Download the JAR file and run it locally with `java -jar agent-fusion.jar --path /path/to/your/codebase`. Wait for the embedding model to process your files (takes 1-5 minutes depending on size). 2. **Query Your Codebase**: Use the prompt template above in your AI agent or terminal. Replace [QUERY] with your specific need (e.g., 'Find all database queries in the user service') and [CONTEXT] with additional details (e.g., 'Focus on MongoDB operations'). 3. **Refine Results**: If results are too broad, add more context to [QUERY] (e.g., 'Only show code using async/await'). For large codebases, use [CONTEXT] to specify directories (e.g., 'Limit search to /src/services'). 4. **Integrate with Agents**: Pipe Agent-Fusion results into your multi-agent system for task automation. For example, use the output to generate pull request descriptions or debug issues without hallucinations. 5. **Update Index**: Re-run Agent-Fusion after major code changes to keep the semantic search accurate. Use `java -jar agent-fusion.jar --update /path/to/updated/files`.
Use Cases
Query your codebase semantically from Claude Code or Cursor without tab-switching or hallucinated APIs
Automatically re-index code and documentation as files change with built-in file watchers
Coordinate design decisions between multiple AI agents with voting and audit trails
Find error handling patterns, database connection logic, and authentication flows across large codebases
Tags
Setup & Installation
Quick Install
No install command available. Check the GitHub repository for manual installation instructions.
Alternative Install (Git Clone)
git clone https://github.com/krokozyab/Agent-FusionRequirements
- Claude Code or compatible AI agent
- Works with: Claude
Quick Start Guide
1
Install the Skill
Copy the install command above and run it in your terminal.
2
Open Your AI Agent
Launch Claude Code, Cursor, or your preferred AI coding agent.
3
Try It Out
Use the prompt template or examples below to test the skill.
4
Customize
Adapt the skill to your specific use case and workflow.
Usage Examples
Prompt Template
Act as an Agent-Fusion local RAG semantic search engine. Query the following codebase for [QUERY] and return only relevant code snippets, documentation excerpts, or configuration files. Use [CONTEXT] to refine the search. If no exact matches exist, provide the closest available references. Format results as: ```[FILE_PATH] [CODE/DOCUMENTATION] ```. Example query: 'Find all API endpoints handling user authentication in the /backend/src directory.'
Example Output
```/backend/src/routes/auth.js
// POST /api/v1/auth/login - Authenticates user credentials
async function login(req, res) {
const { email, password } = req.body;
const user = await User.findOne({ email });
if (!user) return res.status(401).json({ error: 'Invalid credentials' });
const isValid = await bcrypt.compare(password, user.password);
if (!isValid) return res.status(401).json({ error: 'Invalid credentials' });
const token = jwt.sign({ userId: user._id }, process.env.JWT_SECRET);
res.json({ token });
}
// POST /api/v1/auth/refresh - Refreshes JWT token
async function refreshToken(req, res) {
const { token } = req.body;
try {
const decoded = jwt.verify(token, process.env.JWT_SECRET);
const newToken = jwt.sign({ userId: decoded.userId }, process.env.JWT_SECRET);
res.json({ token: newToken });
} catch (err) {
res.status(401).json({ error: 'Invalid token' });
}
}
```
```/backend/docs/authentication.md
# Authentication Flow
The system uses JWT tokens for stateless authentication. Tokens expire after 1 hour and can be refreshed via `/api/v1/auth/refresh`.
## Security Considerations
- Always use HTTPS in production
- Store JWT_SECRET in environment variables
- Rate limit `/api/v1/auth/login` to prevent brute force attacks
```
```/backend/config/auth.config.js
module.exports = {
jwt: {
secret: process.env.JWT_SECRET || 'default-secret-change-in-production',
expiresIn: '1h',
refreshExpiresIn: '7d'
}
};
```Apply to these tools
Browse all toolsJason AI SDR Agent
AI sales agent for lead generation and follow-up
Burp Ai Agent
AI-assisted web application security testing
Microsoft Teams
Get more done every day with Microsoft Teams – powered by AI
GPT for work
Automate your spreadsheet tasks with AI power
Respell
Agentic AI Workflow platform
Notion
Connected workspace for docs, wikis, and projects
Compatible MCP servers
Browse all MCP servers
Find the right skills for your stack
Take a free 3-minute scan and get personalized AI skill recommendations.
Take free scan