Qorven is the open-source, self-hosted alternative to fragmented AI assistants. Install once, own everything, and give each of your Qors — your specialised agents — the channels, memory, and tools to actually get work done.

​
Pick your start

Install in 60 seconds

One curl command sets up Postgres, TLS, and the web UI behind HTTPS.

Have a first chat

Meet Prime, your chief-of-staff Qor, and watch it delegate.

How it's built

The 5-minute tour: agents, channels, memory, tools, gateway.

Channels

Telegram, WhatsApp, Slack, Discord, Email — 20+ surfaces.

​
What makes Qorven different

One Qor, one chat

No more juggling threads. A Qor holds one chat per user, merged across every channel you message it on.

Memory that remembers

pgvector + knowledge graph + periodic dreaming. Compaction turns old turns into searchable facts.

Zero telemetry

No phone-home, per-install encryption key, TLS on by default. The only network calls you didn’t authorise are zero.

​
Who is this for

Self-host on a $5 VPS or a mini PC. One curl | sh install. Your Qor handles email, research, code, scheduling — all through one chat on the channel of your choice.
Multi-tenant Postgres with row-level security. Each teammate has their own Qors; shared rooms for joint work. Free forever for self-hosted.
On-prem deploy, SSO, per-install encryption, audit log, prompt-injection defence, PII redaction, approval-gated destructive tools. Run it air-gapped if you want.

​
Stack at a glance

1

Go backend, Next.js frontend

One binary (~160 MB) with the web UI embedded via go:embed. No Node runtime needed at deploy time.
2

PostgreSQL + pgvector

Sessions, memories, audit log, tenant data — all in one Postgres. Embeddings live next to metadata, no separate vector DB.
3

TLS by default

Local CA generated at first boot. sudo qorven tls install-ca → browsers show the green lock on localhost + every LAN IP.
4

Your choice of LLM

OpenAI, Anthropic, Bedrock, DeepSeek, Gemini, Groq, Mistral, Ollama, LM Studio. BYO keys, rotate automatically on rate-limits, fall back across providers.

​
Licensing

Qorven is released under FSL‑1.1‑ALv2. Free for everyone, including commercial use, with one restriction: you can’t take Qorven and sell it as a competing hosted service.
For AI agents reading these docs: there’s an agent guide with the canonical markdown endpoints, llms.txt index, and the docs_search tool your Qor can call to consult these pages at runtime.