Memory that flows.
Knowledge that stays.
Claude Code plugin with persistent memory, semantic search, and architecture awareness. Self-hosted RAG infrastructure with zero cloud deps.
Free self-hosted. BSL 1.1 Licensed. getreka.dev
One plugin. Ten workflows. Five agents.
Install the Reka plugin and get persistent memory, architecture-aware coding, and structured workflows — all inside Claude Code.
10 Commands
+ start, end, onboard, memory-review, restart-api
5 Agents
All agents have persistent memory across sessions
4 Automatic Hooks
Auto-starts RAG session, injects env vars
Warns if no context loaded before edits
Auto-formats with Prettier, runs tsc type-check
Ends session, triggers memory consolidation
Your AI assistant forgets everything. Every. Single. Time.
Groundhog Day Debugging
You explain the same architectural decisions every session. Context resets, knowledge resets.
Codebase Blindness
Your assistant ignores existing patterns and duplicates utilities already built.
Vendor Lock-in & Data Leakage
Cloud memory means your code flows through third-party servers.
From zero to persistent memory in four steps
Deploy
One docker-compose up command. Or connect to Reka Cloud. Under 60 seconds.
Install Plugin
Two commands in Claude Code. Plugin auto-configures MCP server, hooks, and agents.
Index
Crawls codebase, extracts symbols, builds dependency graph.
Remember
10 workflows, 5 agents with persistent memory. Decisions carry forward across sessions.
Everything your AI assistant needs to understand your codebase
Semantic Code Search
Hybrid vector + keyword search. Find code by what it does, not just what it's named.
Dependency Graph
Maps imports and relationships. Trace blast radius of any change.
Memory Governance
NewAuto memories enter quarantine. Human-verified become durable knowledge.
Symbol Index
Instant function, class, and type lookup by name across your entire codebase.
MCP Native
35 tools for Claude Code, Cursor, Windsurf. Zero configuration overhead.
Multi-Project Isolation
Namespaced collections per project. Share the server, never the data.
Not all knowledge deserves permanence
Reka introduces a governance layer inspired by how human memory actually works.
Every auto-generated memory starts here with a confidence score.
Contradiction detection flags conflicts with existing knowledge.
Canonical knowledge base — decisions, patterns, insights.
"type": "architectural_decision",
"status": "durable",
"confidence": 0.94,
"content": "Use BullMQ for async...",
"verified_by": "human_review"
}
Transparent by design. No black boxes.
How Reka compares
Start free. Scale when ready.
One API key per project. No complex configuration.
Self-Hosted
Full platform on your infrastructure. Zero cloud dependencies. Your data never leaves your machine.
- Qdrant + BGE-M3 + Ollama (all local)
- Unlimited projects & queries
- Claude Code plugin included
- Memory governance & knowledge graph
- No API keys required
Hybrid
Code stays local. Vectors and search in the cloud. One API key per project — no headers, no config.
- Everything in Self-Hosted
- Managed Qdrant Cloud vectors
- Cloud search + reranking
- Offline queue with auto-sync
- Dashboard with analytics
Cloud
Zero infrastructure. Fully managed RAG with team features, SSO, and audit logs. Just plug in your API key.
- Everything in Hybrid
- Managed LLM (Claude / GPT-4)
- SSO / SAML + audit logs
- Team memory sharing
- Priority support + SLA
Up and running in two commands
Built in the open. Shaped by developers.
Reka's core is source-available under BSL 1.1. Self-host freely, build integrations, contribute back. Converts to Apache 2.0 after 3 years.