30 Experts that advise.
6 Agents that act.

Experts are 30 chat-driven specialists (HR, Sales, SEO, Legal, Finance and more). They have their own system prompts and tool catalogues, but they only respond to chat — they never take real-world action on their own. Agents are 6 (and growing) production workers that run on schedules (every 30 min, every 6 h, etc.), call real tools (ollagraph scrapes, WHOIS, AEO audits), and write to a workspace. Every action that could leave the platform — sending an email, publishing a post — is queued for your approval. You pick which Agents are read-only (advisory) vs which can write drafts. See /agents for the full list.

Two things. First, 30 Experts with their own system prompts, files + site grounded into every chat, memory across sessions, and real tool use backed by 200M owned proxies and 4,000+ scraping actors via ollagraph. Second, 6 production Agents that run on schedules and take real actions with approval gates — outbound SDR, SEO audits, content drafting, inbox triage, pipeline coaching, deep research. Generic chat tools rent their web access and don't have the scheduled-action layer at all.

Read-only Agents (SEO Auditor, Deep Research, Pipeline Curator) only produce reports — there's no risk surface. Action-taking Agents (Outbound SDR, Content Publisher, Inbox Triage) write to a workspace that's connected to real channels with your credentials. We gate those behind paid plans so we can offer per-workspace audit logging, approval policies, and the support layer customers actually need for action automation.

Every word in the model's output is counted against your monthly quota, including Agent reports and chat variants you reject. We track per-generation in the History panel so you can audit. Agent tool calls (scrapes, WHOIS) are unlimited — same as Expert tools.

No. Crawling and file uploads are our infrastructure and are unlimited at every tier. You're only paying for the generation tokens that hit the LLM.

All inference is routed through OpenRouter — never Anthropic. Experts use a mix of OpenAI GPT-4o-mini / GPT-4o / Google Gemini / Mistral Large based on task. Agents pin specific models: outbound-sdr / inbox-triage / pipeline-curator on GPT-4o-mini, content-publisher on GPT-4o, and seo-auditor + deep-research on Tongyi DeepResearch 30B A3B (purpose-built for multi-step tool chains). You don't pick. You don't tune.

No. We send your prompts and selected source markdown to the LLM for inference only. We don't fine-tune on customer data and our vendors' APIs don't train on it either.

Yes. Cancel from your billing dashboard. Your plan stays active until the end of the billing period. Billing is handled by Dodo Payments.

Yes. Designed-in. Workspace isolation, per-tenant encryption, delete-means-gone, and a published DPA on the legal page.