Getting Started with Snipara
Get up and running with Snipara in under 5 minutes. This guide will walk you through creating your first project and connecting your MCP client.
Key Takeaways
- 5-minute setup — Create project, get API key, add MCP config
- 90% cost reduction — From 500K tokens to ~5K relevant tokens
- Works with any LLM — Claude, GPT, Gemini, and more
- No vendor lock-in — We optimize context, you use your own LLM
What is Snipara?
Snipara is a Context Optimization as a Service platform. We do not run an LLM — we optimize and deliver relevant context to your own LLM.
Fastest Way: NPX Setup Recommended
One command installs Snipara MCP + RLM-Runtime and configures everything automatically.
npx create-sniparaPrompts for API key, detects your AI client (Claude Code, Cursor, etc.), creates .mcp.json, and optionally configures execution environments (sandbox/docker/local).
60-Second Setup
{ "mcpServers": { "snipara": { "type": "http", "url": "https://api.snipara.com/mcp/YOUR_PROJECT_SLUG", "headers": { "X-API-Key": "rlm_YOUR_API_KEY" } } }}Use your project slug (e.g., my-project) or ID. Then ask your LLM: "How does authentication work?"
Detailed Setup
Step 1: Create an Account
Sign up for a free Snipara account. You will get 100 free queries per month to try out the service.
Step 2: Create a Project
Once logged in, navigate to the Dashboard and click "New Project". Give your project a name and optional description.
Dashboard - Projects View
Step 3: Upload Your Documentation
You can upload documentation in several ways:
- Manual upload - Upload markdown files directly through the web interface
- API upload - Use our REST API to programmatically add documents
- GitHub sync - Connect your repository for automatic syncing (Pro+ plans)
Step 4: Get Your Credentials
Go to your project's API Keys tab in the dashboard. Here you will find:
- Project ID or Slug - Use either the database ID or the human-readable slug in your MCP URL
- API Key - Click "Create API Key" to generate one
Project Slug (recommended - easier to remember)my-projectProject ID (also works)clxxxxxxxxxxxxxxxxxxxxxxxxxxxAPI Keyrlm_xxxxxxxxxxxxxxxxxxxxxxxxTip
Use your project slug (e.g., my-project) instead of the ID for a cleaner, more memorable MCP URL.
Security Warning
Keep your API key secret. Never commit it to version control or share it publicly.
Step 5: Connect Your MCP Client
Add the MCP configuration shown in the 60-Second Setup above to your client. See our Integration Guide for detailed instructions for Claude Code, Cursor, Continue.dev, and other clients.
Advanced: Near-Infinite Context with RELP
For complex questions that span multiple documentation sections, Snipara offers RELP (Recursive Execution Loop Pipeline) — tools that automatically decompose queries and retrieve context efficiently.
Reduced Hallucination
RELP grounds responses in your actual documentation, providing source-attributed answers.
rlm_decompose— Break complex queries into sub-queries (Pro+)rlm_multi_query— Execute multiple queries with shared budget (Pro+)rlm_plan— Generate execution plans for complex tasks (Team+)
With RELP, handle documentation 100x larger than your context window.
Advanced: RLM Runtime for Complex Operations
For complex multi-step tasks that require code execution, recursive decomposition, or deep context retrieval, use rlm-runtime — a Python CLI and library that combines sandboxed code execution with Snipara's context optimization.
pip install rlm-runtime[snipara]rlm run "Implement user authentication with email verification"What RLM Runtime Does
- Task Decomposition — Automatically breaks complex tasks into sub-tasks
- Context Retrieval — Queries Snipara for relevant documentation per sub-task
- Sandboxed Execution — Runs generated code in isolated environments
- Trajectory Logging — Full execution traces for debugging
Configuration
Create an rlm.toml in your project:
[rlm]model = "claude-sonnet-4-20250514"environment = "local" # or "docker"# Snipara integrationsnipara_api_key = "rlm_your_key_here"snipara_project_slug = "your-project"