Straight answers.

One governed substrate for your business: your customers, revenue, content, and traffic as a single graph where every fact carries provenance, AI writes the configuration, and every answer comes with a receipt.

Yes, by installing the Agency GTM pack: companies, deals, a pipeline board, and follow-up automations. The difference from a CRM product is that the opinion is forkable. Everything the pack ships becomes artifacts you own and can change.

The Content pack gives you posts with editorial states, scheduled publishing, and a public content API your site can render from. Because your content and your site traffic live in one graph, attribution (which posts drive paying customers) is a built-in view.

No, and this is structural, not a policy promise. Agents in Wavetable can read, draft, and propose. External sends require a capability minted by a policy you approved, claim acceptance runs through adjudication, and automations go live only after you approve them with a replay report in hand.

Every answer from Ask carries the exact read-only query that produced it, the row count, and a plan hash. Every automation run has a reproducible trace with its cost. Every sentence in a record's narrative cites the claims behind it. You never have to take an output on faith.

Before an automation goes live, Wavetable runs it against your real event history in shadow mode and shows you the counterfactual: how many times it would have fired, on whom, with the actual drafts. You approve behavior you have seen, not behavior that was described to you.

You do. Your workspace is physically isolated (its own database, storage, and index), you can export it, and erasure genuinely erases via crypto-shredding. Wavetable never pools or benchmarks data across tenants.

No. Model calls run against your data to serve you (drafting artifacts, writing cited memory, answering Ask), and we do not use your data to train models, nor permit our model providers to.

Event payloads are envelope-encrypted with per-subject keys. Erasing a person shreds their keys, redacts their claims, and rebuilds projections. The audit log stays append-only; the erased data becomes permanently unreadable.

Yes. Every workspace exposes a policy-scoped MCP endpoint with OAuth 2.1. Claude, a partner's agent, or your own scripts can query the graph, read cited memory, and propose work, within scopes you granted, with every call logged.

Gmail and Google Workspace, Stripe, a first-party site script, Google Calendar, Twilio SMS, and CSV import. Outbound email goes through verified sender identities via Gmail or Resend.

Export a CSV from your current tool and use the import wizard: upload, confirm the column mapping, and rows arrive as claims with the import as their source. Identity resolution then reconciles imported people with the ones already assembled from email and payments.

A modest base fee plus metered usage: events ingested, automation runs, AI tokens, external sends, and MCP calls. Seats are deliberately a minor dimension. There is a free tier for real evaluation, and usage is auditable inside the product.

Connector-derived content is data, never instructions. Message bodies and page content are wrapped as untrusted before any model sees them, and capability checks live in the runtime, not in prompts. A malicious email cannot instruct your agent.

Not yet. The system is designed against SOC 2 control objectives (access logging, change management through approvable artifacts), and we state the current posture plainly on the security page rather than implying otherwise.

Your data stays yours: export is available, and we retain your workspace per the retention terms before deletion. Automations pause; nothing is held hostage.