AtlasClaw Enterprise multi-user AI agent framework
EN 中文
GitHub
Open Source AI Agent Framework for Enterprise

Open Source Enterprise AI Agent Framework

Built for multi-user AI agent deployment, governed execution, and real enterprise AI application delivery.

AtlasClaw helps enterprises move beyond personal agents with SSO-based access, RBAC permission governance, Skill-based system integration, provider-based integrations, webhook AI integration, and enterprise-approved model providers.

Start with the core repo Read the architecture
Enterprise fit Open source
AtlasClaw icon
AtlasClaw

Built for multi-user AI agent deployment, governed execution, and real enterprise AI application delivery.

Multi-user AI agent One deployment for organization users Run one open source multi-user AI agent framework instead of per-user agent stacks.
Systems by Skills Webhook and provider-based integration Use webhook entrypoints and skills to bring AI capability into observability, ITSM, OA, CRM, and other existing enterprise systems.
Enterprise security and compliance SSO, RBAC, and admin control SSO and RBAC keep governed AI agent execution aligned with enterprise-approved LLM providers.
Multi channels IM to Agent to Systems Use IM as the anywhere access layer: users reach AtlasClaw through Slack, WhatsApp, Feishu, WeCom, or DingTalk, then the agent executes skills and interacts with enterprise systems.
Why AtlasClaw for enterprise

Personal agents solve personal workflows. AtlasClaw solves enterprise ones.

Personal agents are useful for individuals. AtlasClaw is built for organizations that need a multi-user AI agent, governed execution, provider-based integrations, webhook AI integration, and model control.

Multi-user deployment

Personal agent

Each user configures their own agent stack, tools, and access separately.

AtlasClaw

One multi-user AI agent deployment can serve users across the organization.

Controlled execution

Personal agent

Unsafe or over-permissive execution is hard to standardize and govern.

AtlasClaw

Governed AI agent execution stays aligned with enterprise SSO and RBAC controls.

Enterprise system access

Personal agent

Individual tools rarely map cleanly onto internal enterprise systems.

AtlasClaw

Webhook entrypoints, Skills, and provider-based integrations let observability, ITSM, OA, CRM, SmartCMP, Jira, and other enterprise systems gain AI capability without rewriting them.

Model governance

Personal agent

Model access is often ad hoc and chosen per user.

AtlasClaw

Connect through enterprise-approved LLM providers and controlled deployment rules.

Enterprise application focus

The core enterprise requirements are built in from the start.

AtlasClaw is designed for organizations that want enterprise AI agent capabilities to live inside actual products, workflows, permissions, and system boundaries.

Multi-user AI agent deployment

Run one multi-user AI agent deployment across teams instead of repeating personal setup for every employee.

Permission-controlled execution

Keep execution under authenticated user permissions so unsafe actions are less likely to slip through.

Skill-based multi-system integration

Connect observability, ITSM, OA, CRM, SmartCMP, Jira, and other internal systems through webhook entrypoints, Skills, and Providers so enterprise software can gain AI capability.

Enterprise-approved model access

Route AI requests through enterprise-approved LLM providers instead of ad hoc user-by-user model choices.

Deployment modes

Two deployment shapes. One execution model.

Embedded Agent Mode

Embed AtlasClaw inside an enterprise application and reuse the host system's identity, context, and UI.

Standalone Agent Mode

Run AtlasClaw as a multi-user AI agent layer where users authenticate through SSO and access enterprise systems under governed permissions.

How it works

A clear path from user intent to governed system execution.

The framework stays modular: the core handles orchestration, while providers carry the system-specific contracts that make execution reliable.

  • Channels route requests into the AtlasClaw core.
  • The agent engine selects skills and orchestrates execution.
  • Providers translate work into target-system operations.
  • Enterprise systems remain the source of truth for auth and audit.
AtlasClaw enterprise AI agent architecture with channels, core, providers, and enterprise systems Click to enlarge

Overall architecture

Access channels feed the core, which delegates execution to providers and downstream systems.

AtlasClaw core components for agent orchestration, sessions, tools, and provider registry Click to enlarge

Core components

The reusable core centers on the API layer, agent engine, sessions, tools, and provider registry.

AtlasClaw provider runtime overview for Skills, scripts, auth, and downstream enterprise systems Click to enlarge

Provider runtime

Shows how a provider carries auth, skills, scripts, and downstream system execution.

Start here

Choose the way you want to understand AtlasClaw.

Understand the core architecture

See how channels, core services, skills, and providers become one execution model.

Study the provider model

Learn how AtlasClaw packages integrations into reusable provider boundaries.

Explore current and planned integrations

See what exists today, what serves as reference architecture, and where the ecosystem is going.
Open-source entry

Step into the code and the docs.

The website gives structure. The repositories give implementation detail. Together they form the open-source entrypoint.

Explore the repos