Mcp Server Rabel

What is Mcp Server Rabel?

Mcp Server Rabel (Recidive Active Brain Environment Layer) is a local-first AI memory server that gives AI assistants persistent memory with semantic search, graph relations, and soft pipelines. Inspired by Mem0 and evolved for HumoticaOS, it runs 100% locally with no cloud, no API keys, and no data leaving the machine.

How to use Mcp Server Rabel?

Install via pip: pip install mcp-server-rabel (basic) or pip install mcp-server-rabel[full] (adds vector search and embeddings). Add to Claude CLI with claude mcp add rabel -- python -m mcp_server_rabel. Tools include rabel_add_memory, rabel_search, rabel_add_relation, rabel_get_guidance, and rabel_next_step.

Key features of Mcp Server Rabel

• 100% local, zero cloud dependencies
• Semantic memory storage and retrieval
• Knowledge graph with entity relations
• Soft pipelines for guided step-by-step tasks
• Bilingual support (Dutch and English)
• Graceful degradation: works without vector search
• SQLite storage with optional sqlite-vec

Use cases of Mcp Server Rabel

• Remember facts about users, team members, or projects
• Semantically search past memories using natural language
• Build and query a knowledge graph of entities and relationships
• Guide AI assistants through multi-step processes with soft pipelines

FAQ from Mcp Server Rabel

What dependencies does Mcp Server Rabel require?

With the basic install, only text storage and LIKE-based search are available. The [full] option adds vector search via sqlite-vec and embeddings via Ollama nomic-embed-text. Ollama must be running locally for semantic search.

Does Mcp Server Rabel work without internet?

Yes. It is designed to work entirely offline. No cloud services, tracking, or API keys are needed.

How does search work without vector embeddings?

Without the [full] extras, text search uses SQL LIKE queries. With embeddings, semantic (vector) search becomes available.

Can I add relations between memories?

Yes. The rabel_add_relation tool stores a subject–predicate–object triple, and rabel_get_relations retrieves them, enabling a knowledge graph.

What languages are supported for soft pipelines?

Soft pipeline guidance is available in both English and Dutch, more languages are planned.