Product Introduction
1. What Is FinClaw
FinClaw is an enterprise-grade digital employee platform from Finogeeks — a "manageable, controllable, collaborative, and auditable" enterprise AI agent engine and digital-employee platform.
Built on a 100% cloud-native microservices architecture, FinClaw delivers an Agent collaboration network with continuous self-evolution, dynamic skill updates, and enterprise knowledge accumulation — while guaranteeing AI data security. It enables enterprises to rapidly build a platform of AI digital employees capable of independently completing complex tasks.
FinClaw solves the problem that prevents OpenClaw from scaling inside enterprises: it leverages a multi-tenant system and a high-concurrency scheduling engine to handle massive concurrent calls and complex DAG workflows, while precisely detecting and force-circuit-breaking logical infinite loops and runaway tasks to prevent compute cost overruns. This makes it suitable for on-premises deployment in mid-to-large enterprises.
2. Overall Architecture
FinClaw is organized into four layers, bottom-up:
Infrastructure Foundation
- 100% cloud-native microservices architecture
- Four-layer physical security sandbox isolation
- 16-layer financial-grade defense-in-depth matrix
- Four-dimensional hybrid augmented memory store
Agent Engine Layer
- Private enterprise skill center
- Agent Loop closed-loop execution
- Massive high-concurrency scheduling engine
- Long-thread execution channel
Platform Control Layer
- Native multi-tenant system
- Workflow orchestration
- Network admission and three-state strict control
- Global monitoring and tamper-proof audit
Application Interaction Layer
- Office collaboration (DingTalk / Lark / IM tools)
- Mobile integration (ChatKit SDK)
- Business system integration (OA / ERP / CRM)
- Web conversation entry (with info-tech innovation support)
3. Core Capabilities
Financial-Grade Security
Every call is auditable and traceable. Agent execution is strictly confined within four layers of physical sandboxes, complemented by a 16-layer financial-grade defense system — intent sanitization, information-flow taint tracking, egress leak prevention — ensuring data never leaves the domain.
Unified Management Platform
A "god's-eye" control center with global visibility. AI management is centralized; through the multi-tenant system, each department gets an isolated virtual AI space with independent data and security policies, enabling personalized AI services at organizational scale.
Enterprise Skill Center
Distill enterprise knowledge, expert skills, and experience into reusable capability modules. Move beyond traditional RPA orchestration and low-code limitations to a business-driven AI assembly paradigm — turning tacit knowledge into enterprise digital assets and building tireless, role-agnostic digital employees.
Long-Term Cognitive Memory
Enterprise knowledge accumulation and persistent memory for continuous learning. Breaking past vector-DB limitations, FinClaw combines FTS5 inverted-text search, high-dimensional vector search, RRF hybrid ranking, and a dual-time-axis knowledge graph across time scales — with automatic redundant-memory cleanup to balance LLM performance and cost.
4. Current Release
The current release includes: chat entry (AG-UI Server / Orchestrator), core platform services (Conversation Store / Delivery / Inbox / Push), channel services, document services, plus the ai-infra-rs AI runtime. It already supports the full conversational flow.
5. Core Flows
FinClaw is built on a contract-first microservices architecture and implements core orchestration flows:
- Inbound: User query → intent classification → context assembly → LLM inference → response
6. About This Documentation
This site is the operations documentation for FinClaw, covering on-premises deployment (single-machine Docker and Kubernetes cluster), configuration, day-to-day operations, monitoring and alerting, and troubleshooting. For product trials or business inquiries, see www.finclip.com/finclaw or call +86 0755-86967467.