Appearance
Powered by GoClaw
AgentBrain runs on GoClaw, the enterprise-grade AI agent runtime built for production. GoClaw is the underlying execution engine that powers every agent workflow in AgentBrain. It handles agent scheduling, tool calling, memory management, and secure sandboxing.
What is GoClaw?
GoClaw is the deterministic agent runtime that schedules tools, manages memory, and sandboxes execution. It is self-hostable and MIT-licensed at the core. AgentBrain adds enterprise modules (RBAC, audit, knowledge vault, monitoring) on top.
Why a Separate Runtime?
Most agent platforms tie themselves to one LLM provider or one framework. GoClaw is built as a runtime layer so customers can:
- Swap LLM providers without rewriting agents
- Self-host in air-gapped environments
- Run open-source models on their own GPUs
- Audit and replay every agent loop deterministically
Architecture Layers
flowchart TB P[AgentBrain Platform Layer] --> G[GoClaw Runtime Engine] G --> M[Multi-tenant PostgreSQL] G --> L[8-stage agent loop] G --> X[20+ LLM providers] G --> C[7 messaging channels]
| Layer | Role |
|---|---|
| AgentBrain | Platform layer: UI, RBAC, knowledge vault, observability |
| GoClaw | Runtime engine: scheduling, memory, sandboxing |
Under the Hood
Multi-tenant PostgreSQL
Isolated schemas per tenant. Zero data leakage between workloads.
8-stage Deterministic Agent Loop
Plan → Tool call → Memory update → Reflect, with full replay support. Every state transition is logged so a run can be deterministically replayed.
20+ LLM Providers · 7 Messaging Channels
LLM providers: OpenAI, Anthropic, Gemini, and open-source models including Llama, Qwen, and DeepSeek. Messaging channels: Slack, Microsoft Teams, LINE, Zalo, WhatsApp, web, and REST API.
Self-Hosting GoClaw
You can self-host GoClaw alongside AgentBrain for full sovereign deployment. The enterprise team helps plan the deployment topology for high availability, key management, and provider routing.
For self-hosting at scale, contact enterprise sales with topic "enterprise".
Related
- Architecture for the deeper view
- Platform Overview for the layer above
- Deployment Options