MarketFive Studio

Autonomous AI that researches, checks, calls, and closes deals while you sleep.It doesn't forget to check your content status and record the voice and extract pdfs from calls.

Comment 1

MarketFive Studio

Project Description

Inspiration

We watched sales teams drown in tab switching. One tool for lead research, another for scraping Reddit, another for voice calls, another for CRM sync, and five more for content. By the time a rep stitched together a single personalized email, the lead had gone cold. We wanted one brain that could research a market, feel its social pulse, write the outreach, make the call, and log everything—without the user ever leaving the canvas.

What it does

MarketFive Studio is an autonomous sales intelligence command center built around two brains.

The Research & Intelligence Hub ingests any company or market keyword, then fires parallel reconnaissance across the web. It pulls live funding data and competitor maps through Toolhouse, scrapes news and landing pages via Firecrawl, and harvests raw social sentiment from Reddit through Composio. Everything distills into a single Opportunity Score radar chart—market timing, competitive moat, team strength, traction velocity, risk level, and momentum—so a rep can decide in ten seconds whether a lead is worth pursuing.

The Communication Command Center turns those insights into action. It drafts hyper-personalized cold emails, SEO blog posts, and tweet threads using OpenAI and GLM, then pushes them to HubSpot or publishes them live to X via Composio. For voice, it orchestrates AI calls through Dogra—our self-hosted voice agent on Coolify—schedules them, records them, transcribes them, and pipes the transcripts straight back into the call log. Research, outreach, voice, and CRM sync live in one room instead of ten.

How we built it

We built the entire frontend and orchestration layer in MeDo, using its internal PostgreSQL database to keep leads, research reports, call logs, and outreach content tightly coupled. We wired six external services through MeDo's HTTP Connector and Secrets manager:

  • Toolhouse handles deep startup intelligence—funding rounds, investor landscapes, and risk signals.
  • Firecrawl tears down company websites and news articles into clean markdown for analysis.
  • Composio gives us native Reddit sentiment scraping and live Twitter/X publishing without fighting OAuth ourselves.
  • OpenAI / GLM generate the outreach copy and summarize transcripts.
  • Dogra runs on our own Coolify instance as the voice backbone for scheduling, calling, and recording.
  • HubSpot receives contacts, deals, and published content through its CRM API.

We pre-seeded the database with realistic research on Stripe, OpenAI, Vapi, Notion, and Figma so the demo feels alive the moment you land on the dashboard.

Challenges we ran into

Our original idea was a visual node-based pipeline builder. Three hours into the hackathon we realized we were building a worse n8n. We scrapped the canvas, killed the drag-and-drop, and pivoted hard to the two-center architecture. That decision saved the project, but it meant rebuilding the core narrative overnight.

The second fight was marrying a self-hosted voice stack (Dogra on Coolify) with a no-code app builder. We could not expose WebSockets through MeDo in time, so we fell back to a webhook bridge—Dogra hits a public MeDo endpoint when a call completes, and the UI updates from the database. It is not real-time streaming, but it is robust enough to demo a live call landing in the log seconds after it ends.

Accomplishments that we're proud of

  • The pivot. Walking away from a broken concept and shipping something focused enough to explain in thirty seconds.
  • Social-to-outreach in one breath. Watching Reddit pain points turn into a drafted tweet, then seeing Composio push it live, feels like magic.
  • Voice that actually leaves a paper trail. Most voice demos are isolated. Ours records, transcribes, summarizes, and files the transcript next to the research that triggered the call.
  • Opportunity Score radar. It sounds like a dashboard widget, but when you see it animate after a three-second search, it turns raw chaos into a decision.

What we learned

Scope discipline is the only feature that matters in a sprint. We cut fund management planners, systematic demo creators, Baidu OCR pipelines, and a calendar scheduler—not because they were bad ideas, but because they were boring to demo. We also learned that orchestrating six APIs through a single HTTP connector is faster than building one bespoke backend, provided you treat every external call as a cacheable, fallible event.

What's next for MarketFive Studio

  • More social ears. LinkedIn company intelligence and Instagram sentiment through Composio so B2C and DTC founders can use the same brain.
  • Live voice streaming. Upgrading the Dograh bridge from webhooks to WebSockets for true real-time transcript display during calls.
  • Collaborative rooms. Shared research workspaces so sales teams can annotate opportunity scores and debate leads inside the app.
  • Mobile companion. A stripped-down app for reps to review call transcripts and approve outreach drafts between meetings.
  • Fund intelligence layer. Re-introducing startup fund management planning and LP pipeline tracking as a premium research module once the core engine is solid.

Built with MeDo, Toolhouse, Composio, Firecrawl, OpenAI, GLM, Dogra, and HubSpot.

Share this project:

Updates