Leading product & engineering @tryconversion | 🍕 connoisseur
San Francisco, CA
Joined January 2016
- Tweets84
- Following103
- Followers103
- Likes274
Started experimenting with an input group. Have been using it for user-defined filters where users can add multiple complex conditions in a column.
Levi retweeted
Leaving Marketo used to mean a 6-month migration and a team of consultants. With @tryconversion’s MCP 2.0, it’s now just a sentence you type into ChatGPT.
Today, MCP 2.0 is officially live for all Conversion customers. Connect it alongside your Marketo, HubSpot, or Salesforce MCP, point Claude or ChatGPT at a campaign or webinar program, and it rebuilds the full thing in Conversion.
What makes it special is that our MCP was built to support every tool call you could think of. Building an email, a landing page, or a form. Running and activating workflows. Pulling data across your entire audience. Adjusting your lead scoring model. Building a report. If you can do it in Conversion, you can do it from Claude at the speed of thought. An MCP that can only read your data is just a fancier export!!
So once you’re in, you can do a lot more than migrate your programs. Here are a few of the workflows customers are already running:
→ Webinar agent: one prompt creates the event in Goldcast, spins up the landing page in GitHub with the Goldcast registration baked in, then builds a 4-step email nurture in Conversion that auto-links to the right page at every step
→ Link agent: every link across your emails and landing pages gets UTM-tagged automatically as it’s built, so attribution in Conversion is clean from the first click
→ Gong + Conversion: a weekly agent pulls the most requested problems from your Gong calls, matches them to the features that solve them, and drafts a new weekly email in Conversion around them
→ Figma + Conversion: an email designer skill that generates hero graphics in Figma and drops them into on-brand emails and landing pages in Conversion, ready to ship
→ Meta/LinkedIn Ads + Conversion: launch a lead gen form natively through Conversion, and use the MCP to send your demand gen team weekly suggestions in Slack on how to reallocate budget
Excited to finally get this into everyone’s hands :)
So excited to share what we're working on @ Labs ;) Long horizon agents are the future of marketing automation.
Today we're announcing @tryconversion Labs, our new AI research initiative, alongside our team's results in optimizing context for long-running agents. Our small team of AI engineers took roughly 70,000 tokens out of every request our agent makes, which cuts about two seconds off every turn.
They've been researching how to optimize agent context, which models best tackle long-horizon agent tasks for marketing use cases, and how to orchestrate agents amongst complex datasets with 100+ potential tool calls triggered by ambiguous queries. Building an audience from a plain-English request used to take frontier models like Claude 45 seconds. Our agent now does it in about 2, with the same accuracy, and it's already live for our customers.
If you're curious, our team @vikramsiva29, Leo Tian, @levibkline ,and @jimfutsu has produced two releases:
(1) One-to-One: our benchmark for how well AI personalizes marketing content
(2) Tool search with Jev: how we pick the right tools for our agent out of 100+
We're especially proud of these releases because the team worked deeply alongside our product team to define how agentic experiences actually touch the lives of our customers. Marketing is one of the hardest use cases for agents as success criteria are qualitative and no two companies model their data the same way.
See more at conversion.ai/labs/. If you're an engineer curious about how technology touches the lives of real enterprise marketing and growth teams, we're hiring.
Attribution pushed my limits. Spent 100+ hours talking to marketing leaders at top B2B companies and thinking through the small things. So excited to see this come to life.
Today, I'm excited to launch @tryconversion Mosaic! Apparently the CEO does the launch video now. Nobody told me this in advance 👇
Noticing a new type of burnout across startups. It's a second order effect of AI: because LLMs handle low-cognitive tasks, a growing share of our day goes to high-cognition work, which is inherently more mentally straining.
The project: Changing our core data model.
Lead engineer: "Don't worry, Claude said it'll be super chill."
There’s a category of software I really admire: tools that quietly become part of your life.
You’ve probably never heard of Dolby On, but it might be my favorite app.
I’ve used it to record my piano improvisations over the years, and to relive those moments in time.
A nice reminder that software doesn’t have to do a lot to create immense value.
Polished UI for field details! Used GPT-6 Astra to help with the little things like histogram animations :)
Added a new "context" popover to our select component so we could show more information about different options.
Been using the /design skill in Claude Code a lot recently for rapid exploration of views and components in our codebase.
Nothing is ever fully polished, but helps me with the cold start problem when redesigning.
Preview of our new filter editor powered by GLM 5.3 Flash.
This query required: scanning 1,000+ fields, 10M+ historical values, and finding the "demo" form asset...
All in <5 seconds 🤯
Wrote more about how we built this in my latest article: xitter.cf/levibkline/status/2095…
