An assistant that has actually read your workspace.

An AI assistant that has actually read your workspace.

Brain lives inside Tenon. It can see the projects, tickets and pages you can see, and nothing else. Ask what's stuck, ask for a briefing, ask it to draft the tickets from the conversation you just had. It runs on the model you choose, including one inside your own network.

Start free Or self-host
Why it matters

Generic AI knows everything except your work.

Copy a ticket into a chatbot, paste the answer back, repeat. That is what "AI for work" mostly looks like today: a very smart intern locked out of the building. The useful version has to be inside, reading the same tickets, pages and conversations your team does.

That raises the obvious question: whose model, and where does the data go? Tenon's answer, on Cloud and self-hosted alike, is that it's your decision. Nothing is hard-coded. A superadmin picks Google Gemini, or any OpenAI-compatible endpoint (OpenAI, Azure OpenAI, or a self-hosted server like Ollama or vLLM), and every AI feature, from the floating assistant to the nightly standup cron, routes through that one configured provider.

Brain also respects permissions. It reads through the same Row Level Security that governs people, so it cannot summarise a project you're not a member of. And if AI isn't configured, features say so honestly instead of showing a canned paragraph dressed up as insight.

What you get

Brain AI, item by item.

Workspace-wide assistant

A floating chat available on every screen, with streaming responses and a typing indicator.

Workspace-aware context

Brain can see projects and tickets, and flags tickets that have sat idle for three or more days.

Morning briefing

An AI-generated daily summary so the first ten minutes of your day are not spent hunting.

Draft ticket cards

Describe a problem in conversation; Brain proposes tickets you can accept with one click.

AI executive summary

A real generated report on the insights dashboard, with proper errors if AI is not configured, never a canned paragraph.

Standup digest

The standup cron job writes a readable digest of everyone's updates instead of a raw list.

Bring your own provider

No AI key is hardcoded. Superadmins choose Google Gemini or any OpenAI-compatible endpoint, including self-hosted models.

One shared AI layer

Chat and cron jobs route through a single edge-function abstraction, so every feature uses the provider you configured.

Bring your own provider

Choose the model. Change your mind later.

Configured once in Admin Settings. Switching providers is a form, not a migration.

Google Gemini

Paste an API key, pick a model. Fast to set up, strong at long context.

OpenAI-compatible cloud

OpenAI, Azure OpenAI, or any hosted provider that speaks the OpenAI API. Base URL, model name, key.

Self-hosted models

Ollama, vLLM, LM Studio, or your own inference cluster. Point Tenon at the base URL and no prompt ever leaves your network.

Admin Settings · AI provider
# Option A: Google Gemini
provider   = gemini
api_key    = AIza…
model      = gemini-2.5-flash

# Option B: OpenAI-compatible / self-hosted
provider   = openai_compatible
base_url   = http://ollama.internal:11434/v1
model      = llama3.1:70b
api_key    = (optional)

One abstraction, every feature

All AI calls go through a shared edge-function module (_shared/ai.ts). The floating assistant, the morning briefing, the executive summary on the insights dashboard and the standup digest cron all use it. Configure once; every feature follows.

  • Streaming responses in the assistant
  • Clear "AI not configured" states, never fake output
  • No vendor lock-in: swap providers without touching data
A Tuesday afternoon with Tenon

A morning briefing instead of a morning of tabs.

8:52. Daniel opens Tenon and the morning briefing is already there: two tickets went stuck overnight on the data-warehouse project, the SOC 2 project closed four items, and a poll in #ops closes at noon. He asks Brain: "What's blocking the warehouse work?" It lists the two tickets, who owns them, and the last comment on each.

He types: "Draft a ticket for Ana to review the schema migration plan by Friday." A draft card appears: title, assignee, due date, priority. He clicks Create. The whole thing took three minutes, and the model that did it runs on a box in his own rack.

Daniel, head of engineering, fintech (self-hosted Ollama)

Questions

Common questions.

Do I have to use AI at all?

No. Tenon works fully without an AI provider. AI features show a clear "not configured" state, and scheduled jobs that depend on AI skip gracefully.

Which providers are supported?

Google Gemini via API key, and any OpenAI-compatible endpoint, which includes OpenAI, Azure OpenAI, and self-hosted servers such as Ollama, vLLM and LM Studio.

Can Brain see everything in the workspace?

No. Brain reads through the current user's permissions. If you cannot see a project, neither can the assistant when it is answering you.

Does Brain change data on its own?

No. It proposes (draft tickets, summaries, reports) and you accept. Nothing is written without a person clicking.

Is my data used to train models?

That depends entirely on the provider you choose and the terms you have with them. With a self-hosted model, no data leaves your infrastructure at all.

Fits together with

The rest of the joint.

Chat & collaboration

Real-time channels and DMs with reactions, attachments, page cards, polls, decision records and async standups.

Explore

Workspace insights

Portfolio health, resource heatmaps, C-suite metrics and an AI executive summary across every project in the organization.

Explore

Projects & work tracking

Kanban boards, timelines, hill charts and analytics, so every project has one honest picture of where it stands.

Explore

Bring every piece of work into one place.

Start free on Tenon Cloud in two minutes, no credit card. Or host it yourself for your departments and clients.

Start free on Tenon CloudSelf-host instead