lexisOS
Setting things up…
Setting things up…
Loading features…
Generation for email, LinkedIn messages, SMS, and workflow content runs on Anthropic — no other model provider drafts your outreach.
An organization can bring its own Anthropic API key, encrypted at rest, or fall back to a platform-default key under a monthly USD cap an admin sets. Every generation call reserves budget before it runs and finalizes the actual cost after, so a spending cap holds even under concurrent use — not a best-effort estimate checked after the fact. Every system prompt behind these drafts is stored in the database and versioned on every change, not hidden in application code.
| System | Model | What it does | Precondition |
|---|---|---|---|
| Drafting — email, LinkedIn, SMS, workflow content | Anthropic (Claude) | Writes the message or step content a workflow uses | An Anthropic key is required — bring your own or use the platform key |
| In-app assistant, everyday chat | A self-hosted Qwen2.5-14B model, run locally | Reads your workspace to help you — read-only | Off by default; enabled per deployment |
| In-app assistant, a requested change | Escalates to Claude | Drafts the exact proposed action and its arguments | Nothing runs until you approve — the same gate as a workflow send |
For a change to your CRM data, in plain language.
A bounded read-verify pass checks the request against your real data.
The precise tool call and its arguments — not a vague intent.
Nothing executes until you click approve — the same gate a workflow send goes through.
“AI drafts it. You decide if it goes out.”
Start free. AI drafting stays off until you connect an Anthropic key or turn on the platform key with a spending cap.