The short version
The API and MCP server are public and read-only. You can use nearly all of it without an account, and without telling us who you are. We do not sell data, we do not run advertising, and we do not build profiles of individuals.
What we do keep is operational: enough to run the service, bill the accounts that have one, and know which parts of it are actually used.
Requests to the API and MCP server
Every request is logged for operational purposes. That record includes:
- The route or tool called, response status, and timing.
- A salted SHA-256 hash of your IP address — never the address itself. It exists so “how many distinct callers” is answerable without identifying any of them.
- Your client’s self-reported name and version, from the User-Agent or the MCP handshake.
- For MCP tool calls, the tool name and arguments, and any free-text context your agent chose to send describing why it was calling. That text is written by your agent, not by us — if it puts something sensitive there, we receive it.
Cloudflare, which serves every request, keeps its own edge logs independently of this.
If you sign in
Signing in with GitHub stores your GitHub user ID, your login, and a tier. We ask GitHub only for read:user — we never see your password, your email is not requested, and we cannot act on your GitHub account. You are shown which client is asking before the flow begins, and you can revoke the grant at any time.
Accounts with an API key also accumulate per-day request and quota counters, which is what a usage dashboard and any bill are computed from.
The credential store
If you register a credential for a third-party subnet API, it is encrypted with AES-256-GCM before storage and is readable only by your own account. It is sent to the subnet surface you registered it against and to nowhere else.
Stored credentials expire automatically: after 30 days by default, and at most 90 days if you ask for longer. You can delete one at any time. The store exists so secrets stop travelling through tool arguments, client logs, and conversation transcripts.
What we do not collect
- Raw IP addresses in our own analytics.
- Advertising, cross-site, or third-party tracking identifiers.
- Private keys, seed phrases, or wallet secrets. Wallet signing is non-custodial and happens in your own wallet — never on our servers.
- Payment card details. We do not process card payments.
Retention
- Product analytics are retained by PostHog under its own retention policy.
- Stored credentials expire on the schedule above, enforced by the storage layer rather than by a cleanup job.
- Account records and usage counters are kept while the account exists.
- Chain data — blocks, extrinsics, balances — is public information read from the Bittensor network. It is not personal data we collected, and we cannot delete it from the chain.
Who else processes it
We use a small number of infrastructure providers, each processing data only to deliver the service:
- Cloudflare — serving, edge caching, storage, and DDoS protection.
- Neon — the Postgres database behind accounts and indexed chain data.
- PostHog — product analytics and error tracking.
- Unkey — API key issuance and verification.
- GitHub — sign-in, when you choose to use it.
Your choices
- Use the API and MCP server anonymously — most of it needs no account at all.
- Omit the optional context argument on MCP tool calls if you would rather not send it.
- Delete a stored credential, or let it expire.
- Ask us to delete your account and its records, using the contact below.
Changes
This policy is versioned in the same repository as the code it describes, so any change to it is visible in the commit history alongside the change that prompted it.
Contact
Questions, corrections, or a deletion request: open an issue, or see security.txt for a security contact.