@persistadev/mcp-server

Give your AI persistent memory across sessions

Persista is an MCP (Model Context Protocol) server that provides AI assistants with persistent memory, context management, and intelligent learning capabilities.

Features

🧠 Memory - Save and retrieve user preferences, facts, and behaviors across sessions
📝 Context - Smart context packing with relevance scoring and threading
🔄 Sync - Multi-instance coordination with intent broadcasting
📚 Knowledge - Version-aware best practices and deprecation warnings
🧬 Codebase DNA - Automatic detection of patterns, conventions, and architecture
⚠️ Error Learning - Captures build errors and anti-patterns to avoid repeating mistakes

Quick Start

1. Get an API Key

Sign up at persista.dev and create an API key from the dashboard.

2. Add to Cursor

Add this to your ~/.cursor/mcp.json:

{
  "mcpServers": {
    "persista": {
      "command": "npx",
      "args": ["-y", "@persistadev/mcp-server"],
      "env": {
        "PERSISTA_API_KEY": "pst_your_api_key_here"
      }
    }
  }
}

3. Restart Cursor

That's it! Your AI now has persistent memory.

What It Does

Once installed, the AI can:

• Remember your preferences, coding style, and project decisions
• Detect your codebase patterns and conventions automatically
• Learn from errors to avoid making the same mistakes
• Track git changes between sessions
• Coordinate with other AI instances working on the same project

Environment Variables

| Variable | Required | Description | |----------|----------|-------------| | PERSISTA_API_KEY | Yes | Your API key from persista.dev | | PERSISTA_API_URL | No | Custom API URL (defaults to api.persista.dev) | | PERSISTA_USER_ID | No | Custom user ID (defaults to system username) | | PERSISTA_PROJECT_DIR | No | Override project directory detection |

Available Tools

The MCP server provides these tools to the AI:

Memory

• memory_save - Save memories with confidence decay and code attachment
• memory_get - Retrieve all memories with effective confidence scores
• memory_search - Semantic search through memories
• memory_for_file - Get memories attached to specific files

Context

• context_pack - Smart context packing within token budget
• context_relevant - Get most relevant context for conversation
• context_thread_create - Create conversation threads
• context_summary_save - Save conversation summaries

Sync

• sync_discover - Find other AI instances on this project
• sync_intent - Broadcast editing intent to prevent conflicts
• sync_check_conflicts - Check for conflicting edits

Knowledge

• knowledge_check - Version-aware best practices
• knowledge_deprecations - Get deprecation warnings
• knowledge_scan_code - Scan code for deprecated patterns
• knowledge_learn_pattern - Learn custom patterns

DNA (Codebase Analysis)

• dna_analyze - Deep codebase analysis
• dna_conventions - Get coding conventions
• dna_patterns - Get detected patterns
• dna_architecture - Get architecture info
• dna_add_decision - Record tech decisions
• dna_mark_critical - Mark critical paths

Error Learning

• error_capture - Capture errors for learning
• error_patterns - Get known error patterns

Session

• session_refresh - Reload session context
• session_summary - Get full session summary

Resources

The MCP server also provides these auto-injected resources:

• persista://session/context - Full session context (memory, git, DNA)
• persista://user/memory - All user memories
• persista://codebase/dna - Basic codebase analysis
• persista://codebase/dna-deep - Deep codebase DNA
• persista://git/changes - Git changes since last session
• persista://knowledge/relevant - Relevant best practices
• persista://errors/learned - Learned error patterns

Links

• Website: persista.dev
• Dashboard: persista.dev/dashboard
• Docs: persista.dev/docs

License

MIT