AGENTME — AI Agent Platform for Jobs & Freelance

AGENTME is a full-stack web app where users describe goals in natural language and the platform turns them into missions powered by AI and persistent data — covering job opportunity search and analysis, social content (editorial calendar and hook lab), and freelance-oriented flows (leads, proposals).

The product includes a marketing site, authentication, a unified dashboard, plan-based usage limits, in-app notifications, and transactional email.

Many early-career users and freelancers split their time between LinkedIn, social content, and client outreach using disconnected tools — notes, spreadsheets, ad-hoc AI chats — without a clear history or a single place to see what the agent produced and what was saved.

Three core tensions drove the design:

Fragmentation — different needs (jobs, social, freelance) without a shared "mission → result → history" model.

Trust and control — real auth, first-party user data, and actionable feedback when results are thin or something fails.

Product sustainability — heavy LLM usage and scraping require quotas, secure API routes, and coherent database persistence.

A SaaS-style experience built around a unified mission model, with three modes:

Opportunity — job-oriented search with Playwright-assisted scraping and relevance filtering against the user's query.

Social — editorial calendar and a hook lab for post ideas structured for real channels.

Freelancer — copy-oriented flows for leads and proposals, tied to the same mission model.

Every mission persists with its state, action counts, and suggested next steps. Onboarding and a notification center improve both first-run and return visits. The landing page communicates value, surfaces modes and pricing, and routes users to sign up or sign in.

Frontend — Next.js 16 (App Router), React 19, TypeScript, Tailwind CSS 4, consistent UI across marketing and app.

Backend — Route Handlers in the same repo that validate session, enforce limits, and call agent services.

Data & auth — Supabase for users, profiles, missions, and notifications, with SSR via @supabase/ssr.

AI — Anthropic Claude through the server-side SDK; agent logic organized by mode. Secrets never touch the browser.

Email — Resend for transactional mail scoped to valid sessions.

Client data — SWR for dashboard refreshes without full reloads.

Notable choices: strict client/server split for secrets; lazy-loading Playwright to keep dev startup light; mission metadata reused across modes for context prefill.

An end-to-end usable product: sign up, run agents per mode, browse mission history, and receive key emails. The architecture supports adding modes and business rules without rewriting the session or data core.

A single "mission" mental model reduces perceived complexity even when the pipelines underneath differ significantly.

Planned next steps: E2E tests with Playwright, retention analytics, Stripe billing, and shipping the "Money" mode out of coming soon.