VoltSnip - The Single Source of Truth for Agents. A searchable repository of approved snippets and organizational knowledge, curated by teams to keep agent output consistent. Store code, prompts, configs, runbooks, checklists, and templates alongside your design language and preferences.
Explore · API Docs · Report Bug
⚠️ Development Status
This is a personal project in active development. All data sent to the public API is public and unencrypted. Do not use it for production workloads or sensitive code.
Voltsnip stores team-approved snippets and organizational knowledge with semantic search. Instead of rewriting common utilities or asking an LLM to generate the same pattern repeatedly, you can search for existing implementations and standards that have been tested, reviewed, and refined by others.
Core features:
- Semantic search – Find snippets and knowledge by describing what you need, not just keyword matching
- Approved snippets – Curate vetted patterns so every agent uses the same baseline
- Design language & preferences – Encode conventions, architecture decisions, and style rules
- Not just code – Store prompts, configs, runbooks, checklists, and templates
- Community ratings – Vote on snippets to surface the most useful solutions
- Multi-language support – Store and search across Python, JavaScript, Go, and more
- MCP integration – Native support for AI agents via Model Context Protocol
- HTTP API – Use it from any tool or workflow
For developers:
- Quickly find approved implementations of common patterns (retry logic, data validation, parsing utilities)
- Build a team knowledge base that encodes design language and preferences
- Reduce time spent searching Stack Overflow or reading documentation
For AI agents:
- Retrieve approved snippets and org standards instead of generating from scratch
- Save tokens by injecting compact, vetted snippets and references into context
- Enable knowledge sharing between agent sessions or across a team
Research like the PAL (Program-aided Language Models) paper shows that LLMs perform better when they can reference working code. Voltsnip provides a practical memory layer for this approach.
- API documentation: https://voltsnip-api.thetechcruise.com/docs
- MCP endpoint: https://voltsnip-api.thetechcruise.com/mcp
- Clone and configure
git clone https://github.com/vaddisrinivas/voltsnip.git
cd voltsnip/backend
cp .env.example .env- Edit
.envwith your settings. For local Docker usage:
DATABASE_URL=postgresql+asyncpg://user:password@db:5432/voltsnip
- Start services
docker compose up --build- Run database migrations
docker compose exec backend alembic upgrade head- Access the application
- API documentation: http://localhost:8000/docs
- MCP endpoint: http://localhost:8000/mcp
cd backend
cp .env.example .env
# Edit .env with your database settings
uv sync
docker compose up -d db
uv run alembic upgrade head
uv run uvicorn app.main:app --reloadnpx skills add vaddisrinivas/voltsnip/voltsnip-skillThis is the install path that drives skills.sh listing. See docs/skills-sh-listing.md for details.
npx @vaddisrinivas/voltsnip-skill --output ./SKILL.mdThis is a convenience for local copy only (it does not affect skills.sh ranking).
Edit ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"voltsnip": {
"url": "https://voltsnip-api.thetechcruise.com/mcp",
"transport": "http"
}
}
}If your client expects SSE, set "transport": "sse" with the same URL.
Add to your MCP settings file (typically ~/Library/Application Support/Code/User/globalStorage/saoudrizwan.claude-dev/settings/mcpSettings.json):
{
"voltsnip": {
"url": "https://voltsnip-api.thetechcruise.com/mcp",
"transport": "http"
}
}See voltsnip-skill/README.md for detailed integration guides.
Stack:
- Backend: Python 3.12, FastAPI, SQLAlchemy 2.0
- Database: PostgreSQL 16 with pgvector for semantic search
- Embeddings: fastembed (runs locally, no external API calls)
- Interfaces: RESTful HTTP API + MCP server
Search flow:
- Query is converted to an embedding vector using fastembed
- PostgreSQL's pgvector extension finds semantically similar snippets
- Results are ranked by a combination of similarity score, votes, and usage
- Snippets include metadata (language, tags, author) for context
Common development tasks:
- Need a retry decorator with exponential backoff? Search "retry exponential backoff"
- Parsing structured data from text? Find regex patterns others have refined
- Setting up a common config pattern? Retrieve a template someone's already debugged
Agent workflows:
- Agent encounters a known problem and searches for a proven solution
- Agent creates a working implementation and saves it for future reuse
- Different agents (or runs) share knowledge without requiring conversation history
Near-term improvements:
- Automated secret detection to prevent credential leaks
- Code quality checks on submission (linting, basic static analysis)
- Better snippet versioning and forking
Future exploration:
- Private team instances or namespaces
- CLI tool for terminal-based search and submission
- Editor plugins (VS Code, JetBrains)
- GitHub integration for importing from gists or repos
This is a personal project, but contributions are welcome:
- Bug fixes and performance improvements
- Better search relevance or ranking algorithms
- Documentation improvements
- New integrations or tooling
Please keep PRs focused and avoid submitting proprietary or sensitive code to the public instance.
voltsnip/
├── backend/ # FastAPI application and MCP server
├── frontend/ # Web interface (HTML/CSS/JS)
├── voltsnip-skill/ # MCP skill documentation and references
├── scripts/ # Build/deploy scripts
└── local_storage/ # Local dev storage (optional)
MIT License - see LICENSE for details.
Connect: LinkedIn
Note: Significant portions of this project were developed with assistance from AI coding tools.