Privacy policy

TokenX is operated by Hexa. It helps organisations see how their teams use AI agents such as Claude Code, Cowork and Codex, and coaches each person to get more out of them.

An organisation decides to use TokenX and connects its own tools. For the data we process on its behalf, that organisation is the controller and we act as its processor. For your account and this website, we are the controller. Questions about either go to privacy@tokenwisex.com.

• Your account: your name, email address and photo, from Google or Slack when you sign in.
• Your profile: your job title, a short description of your work and your department. You or your admin write them. When your organisation leaves it on, an empty job title and bio are filled in once from the public web (a search on your name and your company); you can edit or clear them.
• AI agent usage: what your organisation’s admin set up for everyone, sent by your agents over OpenTelemetry: sessions, models, tokens, costs, tools, skills and timestamps. Depending on your organisation’s setting, it can include the conversation text. Secrets such as API keys are removed before anything is stored.
• Skills and routines you choose to publish from your assistant. They stay personal unless you share them with your organisation.
• Slack, when your organisation connects it: the workspace, which Slack account is yours (matched by email address), and where our messages to you landed. The app reads a message only when you write to it or mention it in a thread; it does not store those messages.
• Your company’s public website, read to describe the organisation. No member data goes into it.
• Notifications we sent you and how you answered them.

Only to run TokenX for your organisation: the dashboards, profiles and leaderboard, the coaching and notifications, the assistant (on the web, in Slack and in your own AI assistant), and the skill library.

We never sell your data, never use it for advertising, and never use it, or anything we receive from Slack, to train AI models. People who have no TokenX account, or who left the organisation, are never counted or analysed.

Summaries, topics, coaching and the assistant’s answers are written by AI models. Before any text reaches a model, we remove secrets and direct identifiers such as email addresses, links and phone numbers. AI output can be inaccurate: check what matters before acting on it.

• OpenAI models, through Vercel AI Gateway, pinned to data centres in the EU: session summaries, the assistant, coaching, skill reviews and the company description.
• Anthropic Claude, served in the EU with zero data retention: a second opinion on skill reviews.
• TypeSafe, which keeps nothing (zero data retention) and runs in the US: classifies prompts, skills and conversations by kind of work. For organisations that require EU data residency, it only reads the company’s public description.
• Exa, a web search, only for the profile autofill above: its queries carry your name and your company’s name, and are not pinned to a region.

• You see everything about your own usage, including your conversations.
• Your colleagues see your usage figures, the skills you use or share, and your profile, as you see theirs. They never see your conversations.
• Your organisation’s admins see summaries and topics of conversations, never their text, and manage the organisation’s settings.
• Conversation text is readable only by a small number of TokenX staff, for support, debugging or security, after recording a reason. You can see who read yours, and why, in your settings.

Our database and application servers are in London. AI processing runs in the EU, except the prompt tagging and the profile autofill’s web search described above, which are not pinned to a region. Everything travels encrypted (TLS), and the access tokens we hold are encrypted or hashed.

The service providers we rely on: Vercel (hosting, AI Gateway, file storage), Neon (database), Trigger.dev (background jobs), the AI providers above, Resend (email), Sentry (error reporting) and PostHog (product analytics). Conversation text and email addresses are never sent to Sentry or PostHog. Slack, GitHub and Google receive data only when your organisation connects them or you sign in with them.

• Conversation text follows your organisation’s setting: kept, deleted about a week after a conversation was analysed (the default), or never stored at all, in which case it is analysed in memory and only the summary and tags remain.
• Usage figures, summaries and skills are kept while your organisation uses TokenX, and deleted with its account. Your profile is deleted when you leave the organisation, and from then on you are no longer counted or analysed; ask us to delete the rest.
• Notification ideas drafted by AI are deleted after 26 weeks.
• Messages we sent you in Slack stay in your Slack workspace, under its own retention.

You can ask to access, correct, export or delete your data, or object to how it is used. Write to privacy@tokenwisex.com; we answer within 30 days. When the request concerns data we process for your organisation, we handle it with your organisation.

You can also lodge a complaint with your data protection authority.

When this policy changes, we update the date at the top and tell organisations of any significant change. Privacy questions: privacy@tokenwisex.com. Anything else: support@tokenwisex.com or the support page.