ClawMax

Multiagent orchestration platform for OpenClaw

ClawMax provides a web-based platform to manage, monitor, and orchestrate OpenClaw AI agent teams. Deploy team templates, visualize workflow DAGs, track progress, and coordinate agents across your entire ecosystem.

---

🔥 Latest Stable Release: v1.9.3

• 1.9.3 is the chat-archives and validation-regression cleanup release after 1.9.2: archive list/resume correctness, safer current-vs-archived conversation behavior, Templates mobile overflow hardening, Workflows first-load initialization fixes, and stronger on-prem custom-skill import guidance.

🛠 Current RC Line: 1.9.5

• Next validation image: 1.9.5-test-rc1
• 1.9.5 is focused on lightweight performance baselining and customer-facing follow-through around skill import UX.
• This line adds machine-readable perf artifacts for core flows like direct agent chat, workflow load/trigger/progress, rolling history, and optional model-matrix chat samples.
• The Skills page now supports uploading custom-skill ZIP bundles directly, which is the recommended path for cloud, containerized, and on-prem dashboard runtimes.

✨ 1.9.x Highlights

• 1.9.3: chat archive correctness, archived conversation resume/restore cleanup, Templates mobile overflow hardening, Workflows first-load initialization fixes, and clearer on-prem custom-skill import guidance.
• 1.9.2: clearer workflow/log/notification/doctor diagnostics, safer local skill import/runtime-path handling, and more consistent DocHub/file-open payload parsing across dashboard surfaces.
• 1.9.1: major test backfill push, built-in --coverage mode in the wrapper, and broader server-route/runtime regression coverage to establish a measured stabilization baseline.

✨ 1.8.x Highlights

• 1.8.9: stronger chat/runtime stability, waiting-for-input context routing, better workflow failure deep links, Builder/chat-history/WhatsApp fixes, and notification/workspace-switch cleanup.
• 1.8.8: template/workflow hardening around first-apply defaults, shared communication targeting, opt-in recurring assistant automations, and conservative workflow parallelism cleanup.
• 1.8.7: remaining DocHub/file-open hardening, clearer workflow/provider failures, simpler agent actions, workflow-session retry repair, and audited deprecated-model handling across OpenAI, Anthropic, and Gemini.
• 1.8.5 / 1.8.6: DocHub preview and file-resolution upgrades, cleaner provider messaging, notification dedupe/grouping, and lower background churn.
• 1.8.4 / 1.8.3: agent delete-flow fixes plus Add Agent default-model, AI-create model preservation, and provisioning-warning cleanup.

✨ 1.7.x Highlights

• 1.7.9: first-pass Cognee partner integration, curated plugin install/uninstall flows, runtime-secret chat routing, and stronger partner/plugin regression coverage.
• 1.7.8: regression sweep for templates search, bundled helper skills, resend/email flows, workflow session hardening, and notification/chat link correctness.
• 1.7.7 / 1.7.6: cloud/on-prem email and status-delivery fixes plus image/release pipeline stabilization.
• 1.7.5: broader UI/UX cleanup across agent surfaces, notifications, file-open behavior, mobile/dark-light consistency, and AI authoring flows.

For full 1.8.x / 1.7.x details and RC history, see CHANGELOG.md.

---

🌟 Features

Agent Management

• Visual Agent Roster - View all agents with status, tags, and activity
• Agent Creation - Bootstrap new agents from templates or use built-in system agents to generate your agents passing description
• Organization Templates - Import pre-configured agent teams or organizations
• Skills Assignment - Assign 50+ built-in skills to agents or import your own skill and assign
• Tag-based Filtering - Organize agents by role, project, or capability

Communication & Monitoring

• Real-time Chat - Interactive chat with any agent via CLI proxy
• Group Conversations - Multiagent group chats and communities
• Activity Feed - Track agent actions, file changes, and workflow executions
• Status Dashboard - System health, agent count, workspace metrics, and optionally costs and budget via Opik (requires Comet Opik service)
• Budget Controls - Workspace-level budget visibility and enforcement to pause costly execution before spend drifts
• Opik Trace Attribution - Manual chats and workflow runs now stamp workspace, agent/workflow, and real dashboard user identity into trace metadata so shared projects can still be filtered cleanly by user and workspace

Authentication & Keys

• GitHub OAuth Login - GitHub remains the primary general-purpose dashboard login path
• Email OTP Login - Single-user hosted/operator-managed login mode with allowlisted email(s), short-lived codes, and persistent session cookies after verification
• Workspaces Integrations / BYOK - Users can configure hosted and local model providers plus optional integrations for their agents and workflows
• Keys & Secrets Browser Vault - A central browser-local vault can capture reusable provider and partner keys once, then prefill matching template, workflow, skill, and integration inputs
• Browser-Local Runtime Secrets - Templates, workflows, and skills can request browser-local secrets like API keys, event slugs, or export paths without persisting them to the server by default
• Separated Key Policy - Dashboard/system actions use SYSTEM_* keys; user execution prefers BYOK or USER_* keys

Keys & Secrets safety model: - Values in Keys & Secrets are stored in the current browser only. They are meant to centralize capture and reuse across dashboard forms, not to act as a secure remote secrets manager. - ClawMax uses the browser vault to prefill matching inputs for templates, workflows, skills, and visible partner integrations. Users can still override values locally for a specific apply/run/edit flow. - Browser-local values are not a substitute for proper server-side or infrastructure secret management. For production hosted or operator-managed deployments, keep system/runtime secrets in environment variables, secret stores, or your platform’s native secret manager. - If you share a browser profile or machine, treat browser-local vault contents as locally accessible to that profile. Clear or rotate values when changing environments or handing a machine to someone else.

Workflows & DAG

• Workflow DAG Visualization - Interactive dependency graph with parallel lanes and connecting lines
• DAG Execution Engine - Auto-advance pipeline: complete → check deps → trigger next
• Workflow Designer - Create scheduled or manual workflows
• Progress Tracking - Real-time progress bars from agent stdout activity
• Blocker Surfacing - Agents declare blockers, rendered as actionable notifications
• Agent Targeting - Route workflows by tags, groups, or communities
• Execution History - Track workflow runs and results
• Structured Workflow Inputs - Persist kickoff/start input summaries in execution records for better dashboards and traceability
• Cron Scheduling - Set recurring workflows with flexible schedules
• WORKFLOW.md Format - Define workflows as YAML frontmatter + markdown

Templates

• 50+ Organization Templates - Business, technical, personal, science, travel, hobbies, family, events, markets, product research, and launch proposal templates
• 25 Reusable Agent Templates - Leadership, engineering, research, events, testing, market, product, competitive, astronomy, and prototype roles that can be reused independently
• Editable Workspace Variants - Start from built-in agent and organization templates, refine them, and save workspace-local variants without overwriting system templates
• 5-Step Template Wizard - Team Type → Composition → Communication → Workflows → Preview
• Secrets Step for Secure Inputs - Templates with secret or runtime-input requirements now get a browser-local Secrets step during apply
• AI Generate - Describe a team, AI fills all wizard steps
• Local-First Template Feedback - Templates can collect star ratings and short optional feedback locally in the active workspace, and can optionally proxy the same payload to a remote web sink when configured
• Smart Workflow Customization - Dynamic form fields (dropdowns, checkboxes) from template placeholders
• GitHub Coordination - Toggle to add github skills and inject repo instructions into all workflows
• Category Filters - Business, Technical, Personal, Events, Science, Travel, Hobbies, and Family template categories
• Collapsible Template Sections - Collapse or expand Agent / Organization / Workflow sections while browsing
• TEMPLATE.md Format - Lean markdown format with structured body sections
• Import/Export - Download and upload templates and workflows as .md files

Local template feedback for developers: - Template ratings and short feedback are stored in the active workspace at WORKSPACE/SYSTEM/template-feedback.json by default - This is intentional for local dev and OSS use: you can inspect, back up, diff, or clear that file directly - If you set all three remote feedback env vars below, the dashboard will submit feedback to the remote sink instead of the local workspace file: - TEMPLATE_FEEDBACK_REMOTE_URL - TEMPLATE_FEEDBACK_SUMMARY_URL - TEMPLATE_FEEDBACK_TOKEN - The dashboard client still uses the same local /api/templates/.../feedback routes either way; the server decides whether to use local workspace JSON or the configured remote sink

Workspace Visibility

• Shareable Workspace Dashboards - Generate public read-only links for workspace status, workflows, costs, results, and group chats
• Compact Summary Charts - Dense compact-mode agent/workflow/notification visual summaries
• Display Modes - Compact, Standard, and Detail views for different audiences
• Live Theme Toggle - Shared dashboards follow ClawMax light/dark preference with a top-right toggle

Skills Marketplace

• 50+ Built-in Skills - github, slack, web-search, code-review, and more
• Shipables.dev Registry - Search, browse, and install from 1,000+ skills
• Bulk Assignment - Add skills to multiple agents at once
• Custom Skills - Import from local directory or GitHub
• Partner Skills - Surface partner-backed skill references and curated install actions from Workspaces Integrations
• AI Skill Creation - Generate a new skill scaffold from intent, refine it iteratively, and fill missing SKILL.md sections with guidance

Notifications

• Dynamic Blocker UI - Approval buttons, choice pills, input fields, delegation picker
• Agent Actions - Restart, pause agents directly from notifications
• Progress Bars - Workflow progress in notification dropdown
• Search - Filter notifications by title, message, or type

Open Template & Workflow Registries

ClawMax templates and workflows are open source and community-driven:

• Maximilien-ai/templates — 40+ organization templates with TEMPLATE.md spec
• Maximilien-ai/workflows — 18+ workflow definitions with WORKFLOW.md spec

Contribute your own! Submit a PR with your TEMPLATE.md or WORKFLOW.md — help the community build better multiagent teams.

Multi-Workspace Support

• Workspace Switching - Manage multiple isolated agent environments
• Workspace Tagging - Color-code and label workspaces
• Independent Configs - Each workspace has its own agents and settings

---

🚀 Quick Start

Prerequisites

• Node.js 18+ and npm
• Git for cloning the repository

You do not need to pre-install OpenClaw or add model API keys before setup.

• ./setup.sh installs OpenClaw if it is missing
• model/provider keys can be added later in BYOK or Keys & Secrets after the dashboard is running

Automated Installation (Recommended)

Release Installer (tar.gz bootstrap)

If you do not want to clone the repository first, you can install from public release assets instead.

Latest release:

curl -fsSL https://github.com/Maximilien-ai/clawmax/releases/latest/download/install.sh | bash

Pinned release:

curl -fsSL https://github.com/Maximilien-ai/clawmax/releases/latest/download/install.sh | bash -s -- v1.8.0

What it does: - download a versioned clawmax-vX.Y.Z.tar.gz release asset - verify the matching SHA-256 checksum - extract the release bundle into ./clawmax by default - continue into the normal repo setup.sh flow automatically

🎬 Demo Videos

ClawMax currently includes a couple of legacy local backup demo videos for the two most common walkthroughs:

• workflow execution: SYSTEM/docs/videos/video1-workflow-execution.mov
• add-agent flow: SYSTEM/docs/videos/video2-add-agent.mov

See SYSTEM/docs/DEMO_VIDEOS.md for the current inventory, gaps, and planned refreshed feature-demo recordings.

You can also bootstrap directly with the checked-in wrapper:

./setup.sh v1.8.0

or choose a custom install directory:

``bash curl -fsSL https://github.com/Maximilien-ai/clawmax/releases/latest/download/install.sh | bash -s -- v1.8.0 --dir /opt/clawmax