Documentation

AgentHub vs Alternatives

This page explains AgentHub's product direction relative to adjacent tools. It is not a claim that every listed AgentHub capability is production-complete today.

Status note

AgentHub Desktop, Web, Hub Server, and Edge Server are actively developed. Feishu/Lark production integration, Remote/Cloud Edge, and full Web + Hub + Edge routing are still in development.

AgentHub vs Alternatives

CapabilityAgentHub directionLibreChatOpenCodeClaude CodeCodex CLI
Primary surfaceDesktop/Web workbench with IM collaboration directionChat UITerminalTerminalTerminal
Multi-agent workflowProfiles, runtime adapters, orchestration, review flowSubagents / assistantsSingle CLI flowSingle CLI flowTask tree / CLI flow
Diff reviewWorkbench diff and approval UXArtifactsTerminal/filesTerminal/filesTerminal/files
Local executionDesktop + Local EdgeServer-side chat appLocal CLILocal CLILocal CLI
Team visibilityHub, audit, IM events in developmentChat historyLocal logs/filesLocal logs/filesRollout/session logs
Runtime modelClaude Code, Codex, OpenCode, custom adapter directionModel/provider pluginsBuilt-in runtimeClaude-specific CLIOpenAI CLI
Self-hosting directionHub + Edge splitDocker/serverCLICLICLI
Public SDK statusAdapter contract shaped; public package not stableApp/plugin ecosystemCLI/tool hooksSDK beta / CLI integrationConfig and CLI integration

AgentHub is being built as a coordination layer around existing agent runtimes, not as a replacement for those runtimes. The design focus is shared workbench state, visible approvals, normalized events, and a Hub/Edge boundary that supports local execution first and team workflows later.

The comparison is intentionally about product shape, not a feature-score race. A CLI can be better for fast personal work. A chat UI can be better for general conversation. AgentHub is useful when the work needs state, review, routing, and audit around multiple agent runtimes.

Decision Guide

NeedBetter directionWhy
Fix one file in a private terminalUse Claude Code, Codex, or OpenCode directlyFast startup; state only needs to live in the terminal
Maintain several agent runtimes over timeAgentHub Desktop + adapter contractProfiles, events, approvals, diffs, and failure states need one vocabulary
Let a team inspect task progress and resultsAgentHub Hub/Web directionSessions, projects, devices, audit, and shared review become necessary
General chat or knowledge-base conversationLibreChat or another chat UIChat UIs are better for provider/model aggregation and general Q&A
Create tasks from Feishu/LarkAgentHub integration direction, currently in developmentIM entry requires binding, card actions, async queues, and Hub tasks

What AgentHub Does Not Replace

  • It does not replace Claude Code, Codex, OpenCode, or similar runtimes. AgentHub uses them through adapters.
  • It does not replace TokenDance Gateway. Model API keys, provider routing, quota, and billing belong to Gateway.
  • It does not replace TokenDance ID. Identity, third-party login, and account linking belong to TokenDance ID.
  • It does not replace GitHub PR review. AgentHub can organize diffs, artifacts, and suggestions, but final review should still use the repository's normal PR flow.

Evidence For Comparisons

If public pages claim AgentHub is better suited for a team-collaboration dimension, that claim should have matching evidence:

DimensionEvidence needed
Desktop workbenchCurrent Desktop screenshot, theme/language sync, mock aligned with real UI
Runtime adaptersEvent examples, error codes, real CLI smoke or conformance evidence
Hub/Web collaborationHub session, device routing, Web review state, audit screenshot
Feishu/LarkWebhook/card callback, queue, binding, failure-state evidence
Remote/Cloud EdgeDevice proof, workspace allowlist, target authorization, relay/provisioning

When To Use AgentHub

AgentHub is the right choice when you want:

  • a Desktop workbench close to the local filesystem and runtime process;
  • a Hub-backed Web surface for collaboration, review, device routing, and audit;
  • multiple runtime adapters under one product vocabulary;
  • future IM entry points where Feishu/Lark can create tasks, show progress, and collect approvals.

Use the original CLI directly when you only need a private terminal session and do not need shared review, routing, or audit.

Do not use AgentHub as the source of truth for provider accounts, social login, or raw model credentials. TokenDance ID owns identity, TokenDance Gateway owns model API keys, and local runtime CLIs keep their own authentication setup.