Imagine the web without open servers.Imagine browsing without an open browser.Imagine AI without open.

Apache and Mozilla made the early internet everyone's to inspect, improve, and run. The assistant should be no different. Commons is the open one — for your group's questions, tools, tasks, and models — with the source readable, every answer on the record, and the whole stack yours to host.

Tools & connectorsTasks & remindersChartsWeb searchAny modelSelf-hosted

Create a group Join with an invite code

01Open source

Trust by inspection, not promise. The code that handles your group's prompts is there to read, audit, and fork. Openness is the only privacy policy you can verify yourself.

02An auditable ledger

Every answer is on the books: the exact model that produced it, what it cost, and why the router chose it. When an agent acts for your group, nothing happens off the record.

03LLM agnostic

Claude-class quality without the Claude-only contract. Run open-weights like Inkling next to Grok and GPT, swap models without losing your history, and never wait on one vendor's roadmap or pricing.

04Local and on-prem

Point Commons at models on your own hardware, or keep the whole stack inside your network. Your group's data doesn't have to leave the building to get good answers.

Why open

Claude and ChatGPT are excellent assistants you have to take on faith. Commons is the same kind of assistant, built in the open — so the code, the costs, and the decisions are yours to inspect, not theirs to describe.

Separate AI seats

~~Typically $20–60~~ /person/mo

→

One shared group

Usage-based

The router picks the cheapest model that can handle each request — and the ledger shows you what that saved.

Commons v0.1 · the open assistant · auditable · model-agnostic · self-hostable