No bashing, no gotchas. Each of these tools is great at what it was built for. This page is so you can pick the right fit — which is sometimes not Qorven.
At a glance
| Tool | Shape | Best for |
|---|---|---|
| Qorven | Self-hosted multi-agent platform | Teams & operators who want a full product with channels + memory + ops |
| ChatGPT Teams | SaaS chat UI + custom GPTs | Individuals who want zero setup and are happy with one vendor |
| LangChain / LangGraph | Python/TS library for agents | Developers building agent features inside their own app |
| AutoGen | Python library for multi-agent conversations | Researchers experimenting with agent patterns |
| CrewAI | Python framework, opinionated | Teams prototyping role-based agent workflows |
| Letta / MemGPT | Hosted memory-first agents | Long-running agents with deep memory, not multi-channel |
| OpenClaw | Similar shape, personal focus | Individual power users, mobile-first |
Qorven vs. ChatGPT (including Custom GPTs + Teams)
| Qorven | ChatGPT Teams | |
|---|---|---|
| Setup | One-line install, runs on your box | Sign up, no install |
| Where data lives | Your Postgres | OpenAI |
| Model choice | Any — you BYO | GPT-4/5 family only |
| Channels | Web + TUI + 20+ integrations | Web + mobile only |
| Multi-agent | Built-in (Prime + specialists) | Manual via Custom GPTs, no delegation |
| Cost | Your LLM bill, no SaaS markup | $30/user/mo + model costs |
| Telemetry | Zero | Standard OpenAI telemetry |
| Customisation | Full source, extensible | Limited to GPT instructions |
Qorven vs. LangChain / LangGraph
| Qorven | LangChain | |
|---|---|---|
| Form factor | Product (run it) | Library (write code with it) |
| Target user | Ops + end users | Developers |
| State | Built-in (Postgres) | You provide |
| Agents | Create via UI | Code |
| Channels | Built-in | Build yourself |
| Memory | Typed + graph + dreaming | You wire it |
| UI | Web + TUI included | None — you build it |
Qorven vs. AutoGen
AutoGen is excellent for multi-agent research — two agents conversing, critiquing, iterating. It’s a library. Qorven’s rooms are inspired by AutoGen’s multi-agent chat but productised: persistence, UI, approvals, audit, safety policies. AutoGen is where you prototype; Qorven is where you ship.Qorven vs. CrewAI
CrewAI is role-based (you define agents with roles + tasks). Conceptually similar to Qorven’s specialists. Differences:- CrewAI = Python library, you define crews in code
- Qorven = self-hosted product, you define Qors in the UI
Qorven vs. Letta (formerly MemGPT)
Letta is memory-first — long-lived agents with sophisticated memory management. That’s also our main investment. Where we differ: Letta is single-agent-focused and hosted. Qorven is multi-agent, self-hosted, with the broader ops surface (channels, connectors, tools, audit). If memory is your only concern → Letta is excellent. If you need multi-agent + multi-channel + on-prem → Qorven.Qorven vs. OpenClaw
We admire OpenClaw — its docs site inspired ours. Both are self-hosted, multi-channel, multi-agent. Where we differ:- OpenClaw emphasises personal use (one user, many channels)
- Qorven supports team use out of the box (multi-tenant Postgres RLS, per-user Qors, team rooms, audit)
- Qorven’s memory system includes the knowledge-graph + dreaming layers
- OpenClaw has deeper iOS/Android native-node integration
When NOT to use Qorven
- You want managed. Use ChatGPT Teams, Claude.ai, or similar SaaS.
- You’re building agent features into your own product. Use LangChain/LangGraph.
- You need a stateless function-calling library. That’s OpenAI’s raw SDK.
- You’re fine with one channel (web only) and one vendor. ChatGPT or Claude.ai.
Where next
Architecture overview
Decide from what we actually are.
Licensing
Is Qorven usable for your case.