FAQ

Short answers to the questions we hear most, including the ones where the honest answer is no. When a limit exists, we would rather you read it here than discover it later.

Is Torno a replacement for my analytics, ads, and payments tools?

Partly, and deliberately. Torno bundles the instruments the Loop needs — first-party events, funnels and retention, Experiments, revenue reporting, ad operations — so one evidence model spans them. It does not claim to beat every specialist tool at every standalone job, and payment and ad-auction rails stay provider-owned: Stripe still bills, Google and Meta still run auctions. See Torno vs the stack.

Does Torno guarantee growth or causal certainty?

No. Readouts use a closed evidence vocabulary — insufficient_evidence, trending, significant_win, significant_loss, guardrail_breach — with bounds and a data basis; no AI silently declares a winner, and Decisions are recorded separately with rationale. A Change with no declared exposure is explicitly unmeasurable and excluded from causal claims, ship-and-watch is labeled as weaker evidence, and ad-platform conversions are labeled separately. See Experiments.

Can agents spend money or publish ads on their own?

Only within policy. Spending actions pass through spend ceilings and reservations, with a default agent cap of $50 per day. Publishing an ad requires human approval under the shipped default policy. Every governed action leaves an audit trail. Details in Policy and approvals and Ads.

How fast can I start?

torno init detects your stack, wires the SDK, and creates a first Loop. The flow is designed so that a first closed Loop takes about 15 minutes — a design target the onboarding is built against, not a benchmark we claim for every stack. See Getting started.

Is Torno GDPR or SOC 2 certified?

Torno does not make certification claims. What it offers are mechanisms: consent enforced at collection, retention windows (raw events default to 12 months), data-subject-request flows, k-anonymity on Cohort reads (default k=25, hard floor 10), and quarantine for untrusted content. How those map to your obligations depends on your jurisdiction. See Identity and privacy.

What does Torno do with content it finds on the public web?

Public-web and voice content is quarantined, provenance-labeled, and fenced as data everywhere a model, agent, or UI touches it; anything derived from it carries an untrusted-derived label. Torno performs no automated outreach to people found through listening. Strangers on the internet are evidence, not a mailing list.

Is there a remote MCP endpoint?

Not yet. MCP is served locally over stdio via torno mcp serve, generated from the same capability map as the API and CLI, and it talks to Torno over the authenticated API. A remote MCP endpoint over HTTP is not currently available. See The agent interface.

How do agents and the CLI authenticate?

Agents use scoped API keys, supplied via TORNO_API_KEY or --api-key; keys are shown once and stored as digests. Humans use torno login, a device-authorization flow. One honest note: human sessions are stored in a mode-0600 file at ~/.config/torno/config.json, not the OS keychain — keychain storage is not implemented yet. Prefer scoped API keys for automation. See The CLI.

Is there a desktop app?

Not as part of v1. A thin desktop wrapper of the web app exists to keep the stack honest, but it is not a supported product surface today. The web app and the CLI cover human and agent use.

Can I export my data?

Yes. The public API includes paginated reads across the full ontology — Loops, Signals, Insights, Hypotheses, Changes, Experiments, Decisions — plus raw-event export and webhooks for continuous delivery. There is no lock-in by omission. See The agent interface.

Does Torno add a network call to my users’ critical path?

No. Assignment runs locally from signed cached bundles, with a sync target of at most 60 seconds and local evaluation after that, and it fails open to control or pass-through. The JS core is budget-gated to stay under 10 KB gzipped. See SDKs and Changes.

What is the relationship to Stella Loop?

They are separate products. Stella Loop is an agent-native project management tool; Torno offers an optional sync where accepted Hypotheses become work items, ship events activate linked Changes, and Results post back as evidence. Outages queue and degrade — the Torno Loop never blocks on Stella. See Stella Loop sync.