What is Beever Atlas?

On May 15, 2026, Hong Kong-based Votee AI and Toronto-based Beever AI open-sourced Beever Atlas — an LLM Knowledge Base that transforms team conversations from Slack, Microsoft Teams, Discord, Telegram, and Mattermost into a structured, searchable memory layer for any AI assistant .

The project responds directly to OpenAI founding member Andrej Karpathy viral call for "an incredible new product" for LLM knowledge bases, moving beyond manual file uploads to chat-native ingestion .

Key Features

  • Chat-Native Ingestion: Automatically ingests conversations from Telegram, Discord, Mattermost, Microsoft Teams, and Slack — no manual uploads
  • Native MCP Server: Cursor, AWS Kiro, Qwen Code, OpenClaw, and Hermes Agent can query team knowledge directly
  • Neo4j Knowledge Graph: Typed entity relationships between people, projects, technologies, and decisions — not just text search
  • Multimodal Intelligence: Text, images, voice, video, and PDFs unified in one searchable memory layer
  • Zero-Install Web UI: No Obsidian or command-line required

Two Editions

  • Open Source (Apache 2.0): Free for individuals — solo developers, researchers, personal knowledge management
  • Enterprise Edition: For teams with permission mirroring, SSO, immutable audit logs, prompt injection defense, and BYOC deployment

Key Differentiators from Karpathy Prototype

  • Chat-native (vs manual file uploads)
  • Multi-user & team-ready (vs single-user)
  • Full Neo4j knowledge graph (vs text-only)
  • Native MCP server for any AI assistant
  • Multi-modal (text, images, voice, video, PDFs)

Enterprise Security Features

  • Permission Mirroring: Mirrors Slack/Teams permissions exactly — if user cannot access channel, AI cannot use that information
  • Immutable Audit Logs: Permanent, tamper-evident record of every action
  • Prompt Injection Defense: Guards against jailbreak attempts
  • BYOC (Bring Your Own Cloud): Runs in customer AWS/Azure account
  • CMEK/BYOK: Customer-managed encryption keys

Pricing

  • Open Source: Free (Apache 2.0)
  • Enterprise: Custom pricing for teams
  • Managed Cloud: Planned for H2 2026

Pros

  • Solves real problem — conversational knowledge loss
  • Open source with permissive Apache 2.0 license
  • Native MCP server for any AI assistant
  • Neo4j knowledge graph enables structured retrieval
  • Enterprise security features for regulated industries

Cons

  • Requires self-hosting for open source version
  • Enterprise features not in open source edition
  • Setup requires Docker and infrastructure knowledge
  • Managed cloud version not yet available

Who Should Use It?

Perfect for: Organizations suffering from conversational knowledge loss, teams using Slack/Teams/Discord extensively, and developers building MCP-native agent memory.

Verdict

Beever Atlas answers Andrej Karpathy call for structured LLM knowledge bases. The combination of chat-native ingestion, Neo4j knowledge graphs, and MCP-native architecture makes it the most thoughtful solution for team conversational memory .

Rating: 4.5/5 - The memory layer every AI agent has been missing.