ClawCrossHub
Local-first AI workspace

Run AI teams locally. Design workflows visually. Share flows through ClawcrossHub.

Clawcross combines a local OpenAI-compatible API, a web UI, Teams and expert personas, OASIS orchestration, OASIS Town, living GraphRAG memory, ClawCross Creator, and a Claude-Code-style WeBot runtime. ClawcrossHub is the flow distribution platform that helps those reusable flows travel.

Discover flows on ClawcrossHub. Run them in Clawcross locally.

OpenAI-compatible API
OASIS workflows
OASIS Town
GraphRAG memory
ClawCross Creator
WeBot subagents
ClawcrossHub icon

Flow distribution platform for Clawcross

Two surfaces, one workflow loop

Clawcross is the runtime and operating model. ClawcrossHub is the distribution surface for reusable flows.

🧩

Clawcross runtime

Local workspace, Teams, personas, orchestration, memory, bots, and operator tooling.

📦

ClawcrossHub distribution

Browse, publish, and share reusable Clawcross flows as portable workflow artifacts.

🔁

One connected loop

Build and run locally in Clawcross, then package and distribute flows through the Hub.

Install via any AI coding assistant

README and SKILL.md are written so Codex, Cursor, Claude Code, and similar assistants can set Clawcross up with the documented flow instead of a hand-written checklist.

Clone https://github.com/ClawCross/ClawCross.git, read AGENTS.md, and install Clawcross.

Media

See it in action

Every visual on this page comes from the actual running system.

Demo Video

A quick walkthrough of the runtime, UI, and orchestration working together.

ClawCross poster

Poster

The core visual carried over from the main ClawCross repository.

Core product

Why Clawcross feels different

The differentiation is not one isolated feature. It is the combination of local control, visual orchestration, live observability, sharable Team artifacts, and an agent-friendly operating model.

👥

Teams and expert personas

Combine internal agents, external APIs, OpenClaw members, personas, and reusable workflows inside one Team.

🔀

OASIS orchestration

Run sequential, parallel, conditional, selector, and DAG-style workflows instead of a single chat thread.

🏘️

OASIS Town + ReportAgent

Watch a live pixel-town and swarm graph, then ask ReportAgent for graph-backed explanations.

🧠

Living GraphRAG memory

Persist topic memory in local SQLite with optional Zep mirroring for external retrieval.

🛠️

ClawCross Creator

Turn a task description or discovered SOP pages into editable roles, personas, and workflows.

🤖

WeBot runtime

Use a Claude-Code-style delegated runtime with profiles, approvals, plans, verifications, and isolated child workspaces.

Evidence

Real product surfaces, not a marketing wishlist

These capabilities already exist in the Clawcross repository today.

🖥️

OpenAI-compatible local API

Expose a local /v1/chat/completions endpoint so standard clients and tools can talk to Clawcross.

📡

Bots, voice, and remote access

Telegram, QQ, TTS/STT, login tokens, and Cloudflare Tunnel push Clawcross beyond a desktop-only tool.

📈

TinyFish monitoring

Track competitor pricing pages, store snapshots, and detect changes over time as part of the operator stack.

📦

Team snapshots and imports

Share Teams, YAML workflows, and snapshots as portable artifacts that can move between environments.

Workflow loop

How the loop works

01

Install and start

Run setup -> configure --init -> start, or hand the repository to an AI coding assistant with SKILL.md.

02

Configure the model

Use the first-login setup wizard to detect models or import settings from OpenClaw or Antigravity.

03

Browse, build, and share flows

Discover flows on ClawcrossHub, then import, adapt, or recreate them inside Clawcross.

Audience

Built for builders, operators, and AI coding agents

🚀

Founders and small teams

Spin up reusable experts for product, engineering, marketing, research, support, and testing.

🧭

Local-first AI operators

Keep the runtime near your machine while still exposing a standard API and optional remote access.

🧑‍💻

AI coding assistants

Clawcross is documented so coding agents can install, operate, and extend it with progressive disclosure.

Prerequisites

Before you start

An LLM provider (OpenAI, Anthropic, etc.) is needed for chat and OASIS orchestration.

OpenClaw integration is entirely optional — you can skip it during setup.

Remote access still requires authentication even when localhost runs passwordless.

Call to action

Start on the Hub. Run it locally in Clawcross.

ClawcrossHub is where flows are discovered and distributed. Clawcross is where the runtime, teams, orchestration, and memory actually execute.