LobeHub logo

LobeHub

★★★★½ 4.5/5
Visit site
Category
Agents
Pricing
Freemium

LobeHub is the current product and brand that grew out of LobeChat. It is no longer positioned only as a polished multi-model chat client. The official description is now a “Chief Agent Operator”: a workspace for creating agents, attaching skills and MCP-compatible plugins, and organizing longer-running work through Pages, Projects, Workspaces, Agent Groups, schedules, and editable memory. Individuals can use LobeHub Cloud with bundled credits and synchronization, while technical users can deploy the community edition and bring their own model providers. A separately negotiated Enterprise edition addresses commercial licensing and management needs.

Quick Verdict

LobeHub is compelling for users who want a refined interface, multiple models, knowledge features, and increasingly ambitious agent collaboration without committing to only a hosted or only a self-hosted route. Cloud removes database, upgrade, and model-billing setup; community deployment increases provider and infrastructure control. Those routes are not automatically identical in features, support, license rights, or data flow. For a strictly self-hosted multi-provider chat service, compare LibreChat. For a single-user desktop document assistant, AnythingLLM is lighter.

Best For

The cloud product fits individual knowledge workers, researchers, writers, and multi-model power users who prefer one account and synchronized history. The community edition fits developers willing to operate a database-backed service, supply provider keys, and customize the application. Teams interested in shared agent workspaces, Pages, project context, schedules, and memory will see more value than users who only ask occasional questions.

Organizations evaluating commercial use should read the current LobeHub Community License and discuss Enterprise when they need private models, centralized user administration, brand customization, commercial licensing, or integration support. Teams building deterministic customer-service transactions should consider Rasa; organizations centered on connected enterprise search should consider Onyx.

Key Features

  • Agent Builder: Users can describe an agent and configure its model, instructions, knowledge, and tools rather than treating every conversation as an unstructured blank chat.
  • Models and modalities: The workspace exposes multiple text, vision, image, and other model capabilities. Cloud usage consumes credits, while community deployments can use provider API keys.
  • Skills and MCP-compatible plugins: Agents can draw from a large tool and skill ecosystem. Marketplace scale does not remove the need to assess the publisher, requested credentials, and action scope.
  • Collaborative work structure: Pages provide a shared writing context; Projects organize work; Workspaces support teams; Agent Groups let specialized agents collaborate or run in parallel.
  • Schedules and memory: Scheduled runs can perform recurring work, while Personal Memory is designed to be structured and editable instead of being an opaque global profile.
  • Self-hosting: The official project provides Docker and lists deployment paths including Vercel and Alibaba Cloud. A durable full-stack installation still needs persistent databases, storage, backups, and upgrades.

Use Cases

An individual can keep research, writing, image understanding, and recurring tasks in one agent workspace instead of moving among provider websites. A team can organize a project, share context in Pages, and let different agents research, draft, or review in parallel. Developers can attach MCP tools and compare models through bring-your-own credentials. Knowledge and file features support project research, but they do not replace data governance. Before sharing an agent or enabling a third-party skill, an operator should determine who can read its context, which credentials it can use, and which model receives the material.

Pricing

EditionOfficial public price on July 15, 2026Included boundary
Cloud Free$0500,000 credits per month, 10 MB file storage, 100 vector entries, unlimited Pages
Cloud Starter$12.90 monthly or $9.90/month billed yearly5,000,000 monthly credits, 1 GB files, 5,000 vector entries, Agent Memory
Cloud Premium$24.90 monthly or $19.90/month billed yearly15,000,000 credits, 2 GB files, 10,000 vector entries, priority email support
Cloud Ultimate$49.90 monthly or $39.90/month billed yearly35,000,000 credits, 4 GB files, 20,000 vector entries, priority chat and email support
Community self-hostingSoftware can be deployed by the userInfrastructure and model APIs are separate; review the current Community License
EnterpriseContact salesCommercial license, branding, user management, self-hosted providers, private models, and custom support

Credits map to input and output tokens differently for each model. The “approximately N messages” figures on the pricing page are estimates, not contractual message quotas; prompt size and generated output can change consumption sharply. Cloud also advertises unlimited requests with bring-your-own provider keys, but those requests create a separate provider bill.

Pros

  • Polished experience that combines multiple models, modalities, agents, and project work.
  • A public free cloud tier and visible paid allowances reduce friction for non-operators.
  • Community deployment supports provider choice and local models.
  • Pages, Projects, Groups, schedules, and editable memory extend the product beyond isolated chats.
  • Enterprise provides an explicit path for commercial licensing and private organizational requirements.

Cons

  • Rapid product evolution makes old LobeChat tutorials and screenshots easy to misapply.
  • Credits are not fixed message units, so long-context usage needs empirical cost testing.
  • The repository currently uses the LobeHub Community License; describing it as Apache-licensed or assuming unrestricted commercial rights would be inaccurate.
  • Self-hosters own database migration, backups, storage, secrets, plugin permissions, security patches, and observability.
  • A self-hosted application can still send prompts and files to cloud models, search services, plugins, MCP servers, or external storage.
  • Cloud and community capabilities should be compared feature by feature rather than assumed to match.

Alternatives

ProductBetter whenMain difference from LobeHub
LibreChatSelf-hosted provider access and team chat administration lead the requirementsMore operator-centric configuration; LobeHub offers a fuller official cloud and visual workspace
AnythingLLMDesktop or Docker private-document RAG is the first taskMore direct document-workspace path; LobeHub has richer agent collaboration and interface design
Open WebUIOllama and local inference are the center of the stackMore locally focused; LobeHub emphasizes cloud options, multimodality, and agent operations
DifyA team needs to build and publish workflow-backed AI applicationsStronger app workflow and API delivery; LobeHub is the end-user agent workspace

FAQ

Are LobeHub and LobeChat separate products?

LobeHub is the current brand and direction descended from LobeChat. When using an older tutorial, check its date, repository reference, deployment architecture, and pricing assumptions against current documentation.

Can Cloud users bring their own API keys?

Yes. The pricing comparison lists bring-your-own provider API keys and unlimited message requests. The provider charges those calls separately from LobeHub subscription credits.

Are Cloud and self-hosted features identical?

Do not assume so. Cloud includes managed operation, synchronization, platform credits, and support tiers. Enterprise has commercial licensing and management capabilities. Compare the current matrix for each required feature.

Does self-hosting keep all data on the server?

Not necessarily. Cloud model APIs, search, MCP tools, plugins, and external storage all create outbound paths. A local data boundary requires local component choices plus network verification.

Are credits the same as tokens?

No. LobeHub maps model input and output tokens to credits at model-specific rates. The same credit balance can therefore support very different workloads.

What should a business verify before using the community edition?

Read the current LobeHub Community License, then identify user-management, branding, private-model, and support needs. Contact Enterprise when commercial licensing or customized support is required rather than relying on an old Apache badge or tutorial.

Bottom Line

LobeHub has moved from an attractive multi-model client toward a workspace where agents are the unit of work. Cloud is the sensible route for convenience and synchronization; community deployment suits teams that can own infrastructure and security. The decisive question is not simply free versus paid. It is who controls and pays for models, who operates data and upgrades, and which license and feature set the organization actually needs.

Last updated: July 15, 2026

Related tools