Agent Desk
Agent Desk is an open-source AI customer support runtime that answers with RAG, runs MCP tools, and hands off to humans.
If you can run Docker Compose, Agent Desk gives you AI triage plus human handoff without a per-seat invoice, and the Apache 2.0 license means no vendor lock-in on the runtime. The v1.6 Agent First rebuild — revisions, run management, MCP tool policies with risk levels — is what separates it from chatbot widgets. It is not plug-and-forget: you own deployment, model endpoints, and upgrades.
Verified 4d ago · liveness 70/100 · cite: rightaichoice.com/tools/agent-desk
- Self-hosting teams that want a free Apache 2.0 support system with AI triage and human handoff
- SMBs automating first-line support while keeping customer data on their own infrastructure
- Developers extending support automation through MCP tools and custom agent skills
- Operations teams that need versioned agents, run history, and ticket follow-up in one system
- Teams without DevOps capacity to run Docker Compose and configure model endpoints
- Organizations that want a fully managed SaaS with vendor-handled uptime and upgrades
- Support orgs needing deep built-in analytics or ready-made Salesforce and Zendesk connectors
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip AgentDesk if you want a vendor to run your support stack for you — deployment is Docker Compose and you own model configuration, upgrades, and uptime.
You pay your model provider directly for every chat, embedding, and rerank call, and that bill scales with conversation volume rather than seats.
As Apache 2.0 software with no license fee, AgentDesk sits below per-seat SaaS support suites such as Zendesk or Intercom on the software line item. The real cost moves to infrastructure and engineering time: you pay your model provider for chat and embedding calls, plus a Docker host and vector storage. For a small team with existing DevOps capacity that beats per-agent pricing; for a team without it, the total cost of ownership can exceed a managed suite.
In short
Agent Desk — Agent Desk is an open-source AI customer support runtime that answers with RAG, runs MCP tools, and hands off to humans. Best for Self-hosting teams that want a free Apache 2.0 support system with AI triage and human handoff, SMBs automating first-line support while keeping customer data on their own infrastructure, Developers extending support automation through MCP tools and custom agent skills. Free to use.
What's new in Agent Desk
Checked 4 days agoAcross the latest 4 updates: 4 changelog entries.
Agent Desk v1.6.3
v1.6.3 refines the embedded support chat window header, status treatment, title/subtitle layout, and action buttons; adjusts Widget iframe corner radii for standard and maximized states.
Agent Desk v1.6.2
v1.6.2 reworks user management drawers, rebuilds the knowledge base dashboard into a list/detail workflow, unifies message rendering, and fixes human handoff transfers.
Agent Desk v1.6.0
v1.6.0 rebuilds the AI Agent runtime around an Agent First architecture, adds Agent revision/publication/run management, MCP tool policies with risk levels, and a FlowGram-based workflow editor.
Agent Desk v1.6.1
v1.6.1 switches the frontend to directory-based static export, fixing 404s on /dashboard/, /dashboard/login/, /support/chat/, and /workbench/ under Docker's default static server.
What people actually say about Agent Desk — is it worth it?
We scanned public community sources for Agent Desk on Jul 3, 2026 and could not establish that the discussion we found is about this tool rather than something else sharing its name. Our own analysis of that scan says the posts were off-subject. Rather than publish a sentiment score built on the wrong subject, we publish nothing here and re-run the scan.
Viability Score
How well maintained and how widely used is Agent Desk? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: October 2026
How we score →Key Features
- Open-source AI customer support system under Apache License 2.0
- Conversation runtime that keeps session state across a conversation
- RAG knowledge retrieval from an enterprise knowledge base
- AI Agent answers first, then escalates when confidence drops
- Human handoff transfers sessions to agents or teams with context
- Ticket lifecycle turns conversations into tracked follow-up work
- Agent revision, publication, and run management (v1.6.0)
- MCP tool management with policies and risk levels (v1.6.0)
- FlowGram-based visual workflow editor (v1.6.0)
- Skill system for custom agent capabilities
- Web widget, direct chat page, and embed SDK for customer chat
- Knowledge base FAQ, documents, chunks, vector indexes, and retrieval logs
- OpenAI-compatible chat, embedding, and rerank model configuration
- Configurable context settings, output limits, timeout, and retry behavior
- Docker Compose deployment
About Agent Desk
Agent Desk is an open-source, AI-first customer support system from Wuhan Huabei Technology that folds live chat, RAG knowledge retrieval, ticket workflows, and human handoff into one runtime. It fits support teams who want AI to take the repetitive first line — password resets, order status, standard procedures — while a person stays one click away for low-confidence or policy-bound threads. Licensing is Apache 2.0, so there's no per-seat invoice and your customer data stays on your own infrastructure. The flow runs end to end. A user enters through the web widget or an open interface; the runtime loads history, AI config, skills, knowledge bases, and tool access; the AI Agent answers from retrieval results plus whatever tools it's allowed to call. Skills and MCP tools let it take real business actions mid-conversation instead of only talking. When it can't finish the job, or the user asks for a person, the session transfers to an agent or team with context intact. Conversations that need follow-up move straight into the ticket lifecycle. Recent engineering has gone into the agent layer. v1.6.0 rebuilt the runtime around an Agent First architecture with revision, publication, and run management, added MCP tool policies with risk levels, and shipped a FlowGram-based visual workflow editor. v1.6.2 reworked user management drawers, rebuilt the knowledge base dashboard as a list/detail workflow, unified message rendering, and fixed human handoff transfers; v1.6.3 refined the embedded chat window header and widget iframe corner radii for mobile and safe-area layouts. Deployment is Docker Compose, models are OpenAI-compatible chat, embedding, and rerank endpoints, and LanceDB is available as the vector store. Against closed SaaS support suites the trade is plain: you get the source, no seat bill, and full data control — you supply the DevOps, model config, and upgrades.
Behind the Verdict
Agent Desk answers a specific question: what does an open-source support stack look like when the AI is a first-class operator rather than a widget bolted onto a help desk? The answer here is a single runtime where the conversation, the knowledge base, the tools the agent may call, the human handoff, and the ticket all share state. We'd reach for this when the team already runs its own infrastructure and the objection to Zendesk-class suites is cost per seat or where the customer data lives. Apache 2.0 plus Docker Compose plus OpenAI-compatible model endpoints means you can keep transcripts and embeddings on your own boxes and swap model providers when pricing or quality shifts. The Agent First rebuild in v1.6.0 is the part worth studying before you commit. Revision, publication, and run management turn agent config into something you can version and roll back instead of hand-editing prompts in production. MCP tool policies with risk levels give you a place to say which external actions the agent may take on its own, which is the difference between a demo and a system a support manager will sign off on. Where it bites: you are the operator. There is no vendor on call at 2am, no SLA, and upgrades land when you pull them. Model configuration — chat, embedding, rerank, context settings, timeout and retry behavior — is yours to tune, and tuning retrieval quality is ongoing work, not a one-time setup. Teams without DevOps capacity will feel that quickly. Compared with closed SaaS suites, you trade managed uptime and ready-made CRM connectors for source access, data control, and no seat pricing. Against a plain chatbot widget, you get the ticket lifecycle and handoff that widgets leave out. If you want the fastest path to a working bot, look at a hosted vendor; if you
Researching Agent Desk? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Real-world workflow fit
Concrete scenarios for the personas Agent Desk actually fits — and what changes day-one when you adopt it.
Stand up AgentDesk via Docker Compose, point model configuration at an OpenAI-compatible chat and embedding endpoint, load the existing FAQ into the knowledge base, and drop the web widget on the help site.
Outcome: The AI Agent answers common queries from maintained knowledge, and low-confidence conversations transfer to a human with context intact rather than starting over.
Configure an AI Agent with model settings, selected knowledge bases, and specific skills, attach MCP tools with policies and risk levels, then version and publish the agent through run management.
Outcome: The agent can take real business actions mid-conversation — order lookups, status checks — under per-tool policy control, and each run is inspectable.
Let the conversation flow into the ticket lifecycle when an issue needs follow-up across time or teams, using the knowledge base and agent config as the grounding layer.
Outcome: First-line traffic is absorbed by AI, and anything unresolved becomes a tracked work item tied back to the original conversation.
Use Cases
- Deploy an AI-first support widget on your website to handle common queries automatically.
- Ground AI answers in your own FAQ and documentation with RAG retrieval.
- Set up human handoff rules for complex or policy-bound conversations.
- Turn resolved chats into tracked ticket follow-ups across teams.
- Design multi-step AI workflows with the FlowGram-based visual editor.
- Let an agent call MCP tools mid-conversation to take real business actions.
- Support customers via WeChat Official Account or WeCom.
- Version, publish, and inspect individual AI Agent runs before putting them in front of customers.
Limitations
- Setup requires Docker Compose knowledge and manual configuration of AI models and integrations.
- The number of pre-built integrations is limited; custom connectors may be needed.
- The project is relatively new, with a small community and thinner documentation.
- No managed cloud offering means you own infrastructure and uptime responsibility.
- Advanced analytics are not included out-of-the-box.
- The UI and some docs are primarily in Chinese, which may be a barrier for English-only teams.
as of 2026-10-04
Verification history
We have re-verified Agent Desk 9 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
Showing the 6 most recent of 9 verification passes.
Free to cite with attribution — this page re-verifies continuously.
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Plans compared
For each published Agent Desk tier: who it actually fits, and what it adds vs. the previous tier. Cross-reference the cost calculator above for projected annual outlay.
Open Source
$0
Ideal for
Self-hosting teams and SMBs with the DevOps capacity to run Docker Compose and configure their own OpenAI-compatible model endpoints.
What this tier adds
Starting tier — Apache 2.0 licensed software at no license cost, including the full conversation runtime, RAG retrieval, AI Agent with skills and MCP tools, human handoff, ticket lifecycle, web widget, and embed SDK.
Where the pricing makes sense
The company stage and team size where Agent Desk's pricing actually pencils out — and where peers do it cheaper.
As Apache 2.0 software with no license fee, AgentDesk sits below per-seat SaaS support suites such as Zendesk or Intercom on the software line item. The real cost moves to infrastructure and engineering time: you pay your model provider for chat and embedding calls, plus a Docker host and vector storage. For a small team with existing DevOps capacity that beats per-agent pricing; for a team without it, the total cost of ownership can exceed a managed suite.
Setup time & first value
How long it actually takes to get something useful out of Agent Desk — broken out by persona, not the marketing-page minute.
A DevOps-capable team with an OpenAI-compatible endpoint and an existing FAQ can typically get a working widget conversation the first day: Docker Compose up, model configuration, knowledge base import, widget embed. Getting MCP tools, skills, and handoff rules production-ready is a multi-day configuration effort, and agent tuning is ongoing. Teams without Docker experience should budget
Switching to or from Agent Desk
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From Zendesk or Intercom: export your help-center articles into the knowledge base, then re-point your site widget at AgentDesk's embed SDK.
- →From a homegrown chatbot script: move FAQ content into the knowledge base and rebuild the reply logic as an AI Agent with skills.
- →From Zendesk or Intercom macros: express standard procedures as agent skills and MCP tool calls instead of canned text replies.
- ↗To a managed SaaS suite: export knowledge base documents and conversation history, then rebuild ticket and handoff rules in the successor's workflow builder.
- ↗To a lighter widget-only tool: keep the knowledge base content and drop the ticket and handoff layers you no longer need.
Integrations
Resources & Guides
Tutorials & Learning
YouTube returned 6 videos for “Agent Desk”, and we withheld 6: 6 did not mention Agent Desk. We are showing none, because we could not prove any of them are about Agent Desk.
Official links
Tools that pair well with Agent Desk
Common stack mates teams adopt alongside Agent Desk, with the specific reason each pairing earns its keep.
Intercom
Intercom pairs Fin, its customer-service AI agent, with a shared-inbox helpdesk so humans and AI work one customer record.
Tidio
Tidio's Lyro AI agent deflects support tickets and hands complex chats to a human help desk
Typewise
Typewise is an AI customer experience platform where AI agents resolve support tickets end to end, set up in plain language with Nova.
Featured Head-to-Head Comparisons
Agent Desk vs Hi Marley
Choose Agent Desk if you need a free, open-source AI customer support platform with full control over workflows and data — and you have the technical ability to self-host. Choose Hi Marley if you are a P&C insurer needing a purpose-built, SMS-first claims communication platform with deep Guidewire and Verisk integrations, even at a premium per-claim cost.
Agent Desk vs Chatlyn
For hospitality businesses needing a purpose-built omnichannel AI platform with PMS integrations and WhatsApp marketing, Chatlyn is the clear choice. However, if you prefer a free, open-source support system that you can customize and self-host, and you don't need hospitality-specific integrations, Agent Desk is more flexible and cost-effective. The choice depends on your industry and deployment preferences.
Agent Desk vs B Rokratt
Choose Agent Desk if you need a customizable, open-source AI support platform with ticket management and human handoff — its latest visual workflow editor (v1.5.3) makes building AI workflows easier. Choose Bürokratt only if you are an Estonian resident or e-resident who needs to access government services via natural language; outside Estonia it has no value. For global customer support teams, Agent Desk is the clear winner.
Alternatives to Agent Desk
View all
Frequently Asked Questions
Categories
Best-of guides
Used Agent Desk? Help shape our editorial sentiment research.