Year
2025 — Present
Role
Founder / Architect
Platform
Trending Society
Stack
TypeScript, Next.js, Cloudflare Workers, Supabase, Trigger.dev
Integrations
20+ Platform APIs
AI Models
Gemini, Claude, GPT-4o, Whisper, FLUX
The Problem
Content teams spend hours manually researching, writing, reformatting, and distributing content across platforms. Every new channel adds another manual step. The tools exist in silos — your CMS doesn't talk to your social scheduler, your analytics don't feed your content strategy, and your AI tools require constant copy-pasting between interfaces.
What Osesame Does
Osesame is a no-code AI agent builder that connects to a production MCP (Model Context Protocol) server with 20+ platform integrations and 350+ tools. Users build custom AI agents that orchestrate real workflows — from content research and generation to multi-channel publishing — all through a conversation interface.
Under the hood, agents call production APIs: querying Supabase databases, triggering Trigger.dev pipelines, publishing to WordPress, managing Shopify storefronts, processing payments through Stripe, sending notifications via Twilio, and generating media with Replicate and ElevenLabs. Every AI call is traced through Langfuse for full observability — token usage, latency, cost, and prompt versioning.
Architecture
The system is built as a monorepo with three layers: a Cloudflare Workers MCP server running at the edge with session management via KV storage, a Trigger.dev task runtime for durable background workflows with retry logic and concurrency control, and Supabase as the data layer with PostgreSQL, real-time subscriptions, and object storage.
Platform integrations are modular — each lives in its own package under @mcp/integrations, with typed clients, Zod-validated schemas, and conditional registration based on available credentials. The registry pattern means adding a new platform is a single file with zero changes to the core server.
Agent Builder Features
Build, customize, and deploy AI agents that orchestrate real production tools — no code required.
Custom AI Agents
Define specialized agents with tailored system prompts, curated tool access, and domain expertise. Invoke any agent by name to handle specific workflows — from content ops to e-commerce management.
Visual Workflows as Tools
Build multi-step workflows visually by chaining LLM reasoning nodes with tool execution nodes. Publish workflows as callable tools — agents can trigger entire pipelines from a single command.
Quick Tool Mentions (@)
Type @ to instantly bind specific tools to a response. Only the mentioned tools are sent to the LLM, reducing token usage and improving accuracy. Create tool presets for common task patterns.
Tool Choice Modes
Control agent autonomy per conversation: Auto (agents call tools freely), Manual (agents request permission before executing), or None (pure reasoning, no tool access). Switch modes on the fly.
Browser Automation
Agents can control web browsers via Playwright MCP integration — navigating pages, clicking elements, filling forms, and extracting data. Multi-step browser tasks execute autonomously.
Realtime Voice Assistant
Voice-powered agent interaction with full MCP tool access. Speak naturally and watch agents execute workflows in real time — from research queries to multi-platform publishing.
Production AI Pipelines
Content Repurposing Agent
Takes any source — social post, article, or video — and generates platform-native content across 6 formats: social captions for 7 platforms, X/Twitter threads, SEO blog drafts, audio/podcast scripts, FAQ schema (JSON-LD), and meta tags.
Single-Article Pipeline (v13)
A 10-step durable workflow that ingests an article URL, deduplicates against the database, scrapes OG images, classifies content by vertical, runs deep web research via Tavily, rewrites with persona-specific Langfuse prompts, generates featured images via Replicate, and publishes to WordPress — all with quality scoring and retry logic.
AI Vision & Media Analysis
Processes images with GPT-4o vision for content descriptions, transcribes videos with Whisper, generates audio summaries with ElevenLabs TTS, and transforms images with Replicate diffusion models. All media assets uploaded to Supabase Storage with CDN distribution.
Social Intelligence Pipeline
Monitors Instagram, TikTok, YouTube, Twitter, Threads, and LinkedIn via Apify scrapers. Normalizes cross-platform data into a unified schema, enriches with AI summaries, and triggers content generation workflows automatically.
Tech Stack
Platform Integrations
- Google Workspace (Gmail, Calendar, Drive)
- Supabase (Database + Storage)
- Shopify (Admin + Storefront APIs)
- Stripe (Payments + Billing)
- Webflow (CMS + Designer)
- WordPress (REST API + WP Automatic)
- Twilio (SMS + Voice)
- Apify (Social Media Scraping)
- Linear (Project Management)
- Airtable (Structured Data)
- Resend (Transactional Email)
- Cloudflare (DNS + KV + Workers)
- OpenAI / Anthropic / Google AI
- Replicate (Image + Video Generation)
- ElevenLabs (Text-to-Speech)
- Inoreader (RSS Intelligence)
- Tavily (Web Research + Extraction)
By the Numbers
350+
MCP Tools
20+
Platform Integrations
7
Social Platforms
6
Content Formats