Steadybase
OpenClaw Origin

What Changed

Feature-by-feature breakdown of how OpenClaw became Steadybase.

What Changed

A detailed feature-by-feature comparison of what Steadybase added, modified, and extended from the OpenClaw foundation.

Feature Transformation

From Single-User Chat to Multi-Agent Platform

Added

Features that did not exist in OpenClaw:

FeatureDescription
8 AI WorkersSpecialized agents with roles, memory, and active workflows
6 Temporal WorkflowsDurable, fault-tolerant multi-step executions
Hierarchical Memory4-scope persistent memory with compression and pinning
Drew Coordinator9-step multi-agent orchestration workflow
Multi-LLM RoutingClaude + GPT-4o + Gemini selection by task type
Human-in-the-LoopSignal-based approval gates with configurable timeouts
Nexus IntegrationCross-namespace Temporal calls
Deal PipelineFull CRM-style deal management
Metrics DashboardToken usage, pipeline health, worker performance
WebSocketReal-time updates for worker status and workflows
VAPI IntegrationVoice call analysis and recording
Slack IntegrationBot notifications and alerts
iOS AppCapacitor-based native mobile client
Rate LimitingTiered rate limits on all endpoints
Audit LoggingSecurity event logging
Role-Based AuthAdmin and viewer roles with invite codes

Modified

Features that existed in OpenClaw and were significantly modified:

FeatureOpenClawSteadybase
ChatSimple LLM chatBrain API with memory context, conversation history, per-user isolation
AuthenticationBasic authJWT with randomized invite codes, role-based access
API StructureSingle route file5 modular API routers (Brain, Workers, Deals, Memory, Metrics)
ServerBasic ExpressExpress with Helmet, CORS, rate limiting, auth middleware
FrontendChat-only UIFull dashboard with workers, pipeline, memory, metrics panels

Kept

Core elements preserved from OpenClaw:

ElementNotes
React SPADashboard builds on the React foundation
Express ServerAPI server extends the Express setup
TypeScriptFull TypeScript codebase maintained
Chat CoreBrain chat feature evolved from OpenClaw's chat
Project StructureDirectory and build configuration preserved

Architectural Shifts

Stateless → Durable

OpenClaw: Every interaction is stateless. Close the browser, lose the context.

Steadybase: Every workflow runs on Temporal Cloud. Close the browser, restart the server, even crash mid-execution — the workflow resumes exactly where it left off.

Single Model → Multi-LLM

OpenClaw: Connected to one LLM provider.

Steadybase: Routes to Claude (reasoning), GPT-4o (content), or Gemini (speed) based on what the task requires.

Chat → Orchestration

OpenClaw: User types a message, gets a response.

Steadybase: User sends a request, Drew decomposes it into subtasks, delegates to specialist agents, waits for human approval, coordinates across namespaces, and delivers a comprehensive result.

Session Memory → Institutional Memory

OpenClaw: Memory exists only during the active session.

Steadybase: Four-scope hierarchical memory persists across sessions, workers, teams, and the entire organization. Workers learn and improve over time.

On this page