AI agent
The agent doesn't assist. It acts. Drafts your contracts. Sends your follow-ups. Logs your time. Chases your overdue invoices. The same one everywhere: web, desktop, WhatsApp, Telegram, Slack.
TL;DR
One agent across web, desktop, WhatsApp, Telegram, and Slack. It writes contracts and invoices, tracks time, sorts your inbox, prepares meetings, and chases overdue invoices on a schedule you set. Memory means you tell it your rates once. It remembers.
What it does
The agent takes action, not just questions. Common ones:
- Drafts contracts, proposals, and invoices from a sentence.
- Reads your inbox. Classifies leads. Drafts replies.
- Logs time. Files receipts. Categorizes expenses.
- Prepares meeting briefs. Summarizes calls. Generates follow-ups.
- Chases overdue invoices on a schedule you set.
- Sends forms, books calls, runs OCR on receipts.
For anything legal or judgment-heavy (a risky clause, a tricky pricing decision, an at-risk project) it pulls in a second opinion from a higher-quality model and explains the trade-offs in plain English.
Where it lives
- Web app. Chat panel on the dashboard. Full streaming, full editor surface, file uploads.
- macOS menubar.Quick chat without leaving whatever app you're working in. Desktop app.
- WhatsApp, Telegram, Slack. Pair once with a link code, then talk to the agent from wherever you already are. Messaging channels.
Same context everywhere
AI inside the editor
Select any text inside a contract or proposal and the AI offers contextual actions: expand a section, simplify legal language, insert a clause, fix grammar, or translate.
Memory and project knowledge
The agent has two kinds of context. They're the reason it gets more useful over time.
Memory (about you)
Things you tell it once and never repeat. Rates. Services. Working style. Who your usual collaborators are. Stored by category and retrieved automatically when relevant.
Project knowledge (about each engagement)
Per-project context: scope, deliverables, deadlines, decisions you made along the way. Auto-populated as you work. You can edit or add anytime.
Proactive nudges
The agent doesn't only respond. It pings you on your last-used channel when something matters: a new email that needs a reply, a meeting in 10 minutes with the brief attached, an invoice 14 days overdue, a deadline approaching, a form submission that looks like a real lead.
Frequency is capped (default 5 per hour, 25 per day) and respects active hours from your profile timezone.
Generating contracts and proposals
Generation uses a scaffold. Pick the engagement shape (fixed price, hourly, retainer, milestone, hybrid) and the jurisdiction (US, UK, EU, AU, CA), then describe the work. A quality judge reads the output and offers a one-click repair if anything important is missing.
You get back a structured retainer contract with your business details, the scope with monthly deliverables, payment terms, jurisdiction-appropriate clauses, a termination section, and a risk score with any missing-clause flags you can one-click fix.
Client-side AI
When clients review a contract or proposal, they get a small AI of their own. It explains sections in plain language, clarifies payment terms, and answers questions about clauses. So they understand what they're signing without you being the helpdesk. Their questions don't use your credits.
Files for context
Drop PDFs, Word docs, images, or text files into the chat. The agent reads them for the current conversation and indexes them for later. Useful for existing contracts, project briefs, screenshots of receipts, or anything you'd rather not retype.
Credits
Legend includes 10,000 credits per month. Different actions cost different amounts:
- Chat message: 10
- Send an email: 25
- Receipt OCR: 50
- Generate an invoice: 50
- Generate a contract or proposal: 150
- Risk analysis: 100
- Meeting brief: 20
- Time entry, expense log: 0
Credits renew on your monthly billing date. Full breakdown on the pricing page.
How to get the most out of it
- Be specific. Project details, budget, timeline up front beat clarifying questions later.
- Build memory early. Tell the agent your rates, services, and how you work on day one. Every future generation gets better.
- Use messaging for quick stuff.Logging time, checking what's due, sending a reminder. WhatsApp one-liners.
- Review before sending. The agent drafts. You decide.
- Iterate.Ask for revisions. "Make this more formal." "Add a confidentiality clause." "Cut the scope section." The agent expects the back and forth.
Use your preferred LLM
Open these docs in ChatGPT, Claude, Gemini, Grok, or Perplexity.
Opens your LLM with this page prefilled as a prompt. The prompt is also copied to your clipboard as a fallback.