Shy R.

Hey folks. Wanted to share a bit of detail around how we're using Claude and Clay together to build really cool things. The workflow is this: Clay is the conductor, Claude is the composer, Tally is the sheet music. 1. We build our live corpus & context layer in Kernel 2. We connect Claude to it via MCP 3. We build versioned prompt specs for enrichment and copy-writing in Claude. With each version, we: --> Push it to github --> Push it to Tally with POST /stash/:id 4. In Clay, we keep an up to date prompt library, with GET /stash/:id running daily 5. In each workbook, we lookup the prompt and feed it into the AI column So it’s always up to date. No more copy-paste. No more forgotten updates. Hours saved. Here's a quick vid of it in action: https://www.loom.com/share/0ed6398636b94afaa225ede84627da8f How are you guys doing this? Feedback is always welcome! Lastly, LMK if you want to try Tally. i'll send any Clay community members a link with a "free-for-life" code.

Hey team, I've had this problem since I started using Clay 2 years ago. How are people doing this? You've got a list of companies or contacts, and you want to get the top 3 accounts in each industry vertical, or region, or the top 5 accounts for each sales rep. I've always found it super clunky and manual. Push rows into different tables and manually limit rows etc. etc. or export out to Excel to do the work. So I created a better way: Tally. How it works is -> Score your rows in Clay (e.g. ICP fit) -> Push the scope and score into Tally -> Get back the rank (+ tiebreaking strategy) -> Data self-destructs if you want it to Demo here. LMK if you want to try, i'll send you community members link with a "Free for life" code.

Hey y'all. Opening up some of the Clay hacks we use with our clients to the broader Clay community at no cost. These are designed to get around some of the most common Clay pains. → RANK — rank within any group of rows, within and across tables. Like a window function for Clay. How we use it: To get the TOP X contacts at a company (or any top X scored rows) → STASH — save and update large JSON blobs, and retrieve them from anywhere. How we use it: to save token-intensive compound research enrichment for lookup, and to push Claude-assisted versioned AI prompts into columns programatically without having to edit cells. → ECHO — native JSON and bypasses the 8kb of text isn't enough. How we use it: to combine multiple text inputs into JSON columns accessible from downstream tables. → RUN — sandboxed JS functions for deterministic responses. How we use it: to trade AI/data credits for action credits, and reliably run things that should be definitive. → RENDER — builds and populates large HTML files from templates How we use it: for ABM and complex dossier generation. DM me or reply in 🧵 if you wanna check it out, I'll send you a link with a free-for-life code attached.