Hong Kong-based Votee AI and its Toronto research lab Beever AI have released Beever Atlas, an open-source LLM knowledge base that automatically transforms team chat logs from Telegram, Discord, Mattermost, Microsoft Teams, and Slack into a structured Neo4j knowledge graph and auto-generated wiki. The tool ships in two editions: an Apache 2.0 Open Source Edition for individuals, and an Enterprise Edition for teams with high-security requirements.
What it does
Beever Atlas ingests chat conversations — text, images, voice, video, and PDFs — and builds a typed knowledge graph with entity relationships between people, projects, technologies, and decisions. The result is a searchable, citation-bearing memory layer that any AI assistant can query via a native MCP server. Supported AI assistants include Cursor, AWS Kiro, Qwen Code, and — coming in Q2 2026 — OpenClaw and Hermes Agent.
Unlike Andrej Karpathy's prototype for an "LLM Knowledge Base," which is single-user, manual, and relies on Obsidian and command-line tools, Beever Atlas is chat-native, multi-user, and requires no manual file uploads or local setup beyond a Docker stack. It runs entirely on-premise with zero telemetry, AES-256-GCM encryption at rest, and private channels filtered by default. Teams bring their own LLM via LiteLLM — either locally through Ollama (Gemma, Qwen, Llama) or through 100+ supported cloud providers.
Enterprise Edition features
The Enterprise Edition extends the open-source core with five capabilities for regulated, multi-tenant environments:
Permission Mirroring — Mirrors Slack and Microsoft Teams permissions exactly. If a user lacks access to a private channel, the AI cannot use information from that channel to answer their questions. Permission changes propagate in under 60 seconds.
Identity & Multi-Tenancy — SSO + SCIM via Okta or Google Workspace. Hard isolation at the database layer prevents data from different companies from mixing.
Audit & Compliance — Immutable, tamper-evident audit logs. Configurable retention for automatic data deletion. Customer-managed encryption keys (CMEK / BYOK).
Trust & Safety — Prompt-injection defense and live evaluations that return "I don't know" with a citation when the model is not confident.
Managed Cloud + Federation — Bring Your Own Cloud (BYOC) on AWS or Azure. Context federation connects to Salesforce, Jira, and BigQuery.
Tradeoffs
Beever Atlas is designed for teams that already use one of the supported chat platforms. If your organization relies on other communication tools, you will need to wait for future integrations. The Open Source Edition is free and self-hostable, but the Enterprise Edition's permission mirroring, SSO, and audit features are not available in the Apache 2.0 version. The OpenClaw and Hermes Agent integration is scheduled for Q2 2026, so users of those tools cannot yet connect to Beever Atlas.
When to use it
Beever Atlas is useful for any team that wants to preserve and search institutional knowledge currently locked in chat conversations. Solo developers and content creators can use the Open Source Edition for personal knowledge management. Banks, government agencies, and large organizations with high-security requirements should evaluate the Enterprise Edition for its permission mirroring, audit logs, and on-premise deployment.
Bottom line
Beever Atlas addresses a real problem — conversational knowledge loss — with a practical, open-source solution that runs on your own infrastructure. The Enterprise Edition's permission mirroring and audit features make it suitable for regulated environments. The tool is available now at github.com/Beever-AI/beever-atlas under the Apache 2.0 license. A managed cloud version is planned for H2 2026.