Questions, answered straight

An always-on AI agent that runs your marketing growth loop — reads results, triages leads, generates creative, optimizes spend, and pursues goals across email, social, and phone — and asks you before doing anything irreversible.

Small businesses that want marketing handled without hiring for it, and agencies running growth loops for multiple clients (one tenant per client).

No. Setup is OAuth clicks and plain-language questions. The most technical thing you'll ever do is optional: edit a markdown file that describes your business.

About five minutes of your input. The agent's first brief usually lands within the following few minutes.

Workflow tools run the steps you drew. Waveguide's agent writes and runs real code in a locked-down sandbox, so it handles situations nobody drew a box for. And the security model is architectural: the model never holds a credential, and irreversible actions require a human — enforced in the network layer, not the prompt.

A persistent goal your agent pursues for days or weeks — "book 10 demos with HVAC companies in Texas." It drafts a brief you approve (targets, channels, sequence, budget), then researches prospects, runs outreach, answers replies, and books meetings, reporting the funnel as it moves. Details on the missions page.

Yes — inside missions. It can place outbound AI calls with consent gates and AI disclosure built in (it always identifies itself as an AI), and attend Google Meets as a voice participant, posting transcripts back to the mission log.

Four autonomy levels per mission: approve every message, approve daily batches, approve the brief and let it send within it, or full autopilot. Whatever you pick, financial and irreversible actions still require a human — that rule has no override.

It's built not to: a tenant-wide suppression list is enforced on every channel, jurisdiction and consent rules gate voice calls, and a hard cross-mission cap means the same person can't be over-contacted even by two missions running at once. Dry-run mode shows you exactly what would go out before anything does.

No. Financial actions are T3 — always human-approved, with no policy override. Ad budget changes are T2 — approved by you or by a standing policy you created, and always inside a hard daily cap that defaults to $0.

Three independent backstops: anomaly detection auto-pauses unusual behavior; the model-budget hard stop (120%) pauses a runaway loop; and the kill switch stops everything instantly, always available in Settings.

It watches for growth-loop-relevant mail (lead replies, platform notifications) using the scopes you approved on Google's own consent screen. Everything it reads and does is in the audit log — inspect it any time.

Untrusted content (email, web, webhooks) is quarantined and typed before it can influence a risky action, and any action shaped by it is forced through approval with a visible taint warning. The worst an injected email can do is ask you a question. Details in Security.

Yes, through a managed cloud browser — for prospect research, competitive pages, and verification. Everything it reads is quarantined like any other untrusted content, and submitting a form requires your approval.

Your tenant is isolated end to end — its own sandbox, its own encryption keys, its own ledger. Our subprocessor list is in the subprocessors page. We never sell or share your data for advertising.

$199, $499, or $1,499/month (pricing) with an included AI budget of $15/$50/$200. Overage at cost × 1.3 with automatic guardrails at 80/100/120% — the 120% stop bounds your worst-case month.

It routes each piece of work to the cheapest model that does it well: a fast model triages signals and reads digests, the flagship model does the real work, and the biggest model handles genuinely hard problems on Growth and Scale plans. Every call is metered to the cent on your Billing page. Running against your own model key is possible in some setups — ask us.

No — ad spend is between you and the ad platforms, controlled by your daily cap. Waveguide bills only for the base plan and the agent's own AI usage.

Billing page → Manage subscription → Cancel. Takes effect at period end; data stays exportable for 30 days.

In your tenant's private wavetable ledger and workspace — isolated storage with per-tenant encryption keys. Nothing is pooled across customers, and nothing is used to train models.

Export everything — workspace, structured data, full logs — as a bundle any time from Settings. Deleting your tenant permanently purges the workspace, credential vault, history, and destroys the sandbox machine. Both are one click, self-serve.

We provide a Data Processing Addendum, a public subprocessor list, self-serve export and deletion, and cookie-free site analytics. Privacy requests: see contact.

Gmail, Google Calendar, Meta Ads, and Stripe as account connections; Slack and Telegram as reporting channels; and tools like Apollo, Hunter, Firecrawl, Instantly, and Mailchimp — or any MCP server you run — via the tool catalog. The integrations page is current.

Yes. Point the agent at any MCP server by URL with OAuth or an API key. Its tools are discovered, risk-classified, and scanned for tool poisoning before the agent can use them; anything suspicious is deactivated until you review it. See the tools docs.

v1 is single-owner per tenant. Approvals can still reach a shared Slack channel where anyone can decide. Team seats are on the roadmap.

One tenant per business. Agencies run one tenant per client (email us about consolidated billing on Scale).

Yes — docs/api. Everything the dashboard does is an API call.

Check spam first. Then request a fresh link from the sign-in page — links are single-use and expire (15 minutes for login links, 24 hours for the welcome link). Still nothing? hello@waveguide.adwave.com.

Starter allows 3 connections; the dashboard will prompt an upgrade for a fourth. Growth and Scale are unlimited.