I spent years doing sales the way most technical founders do: badly and slowly. Every week I’d open Pipedrive, scroll through my pipeline, figure out who needed a follow-up, then spend 3-4 hours writing personalized emails. Each one was different: different deal context, different attachment, different tone depending on where the conversation was.
It worked. But it gobbled up 6-10 hours a week and we had tons of “dropped balls” because the accounting around pipeline management required tons of manual data entry that I hated doing.
So I built TextForge to fix it. And instead of just telling you how it works, I recorded a 3-minute demo showing the whole thing live.
The problem I was solving for myself
Here’s what my workflow used to look like: open Pipedrive, scan my open activities and open deals, identify who’s gone quiet, open Gmail, write a follow-up using one of my pre-saved templates, edit it slightly, attach the right document, send it, log it back in the CRM. Repeat 15-20 times. That’s a Tuesday afternoon gone.
The emails weren’t hard to write individually. But keeping context straight across dozens of deals while cranking out dozens of personalized follow-ups? That’s not hard, it’s just slow. And it’s the kind of slow that makes you put it off until Friday, which means half your deals go a week without a reply.
I wanted a system that could do the research and drafting for me while keeping me in the loop on what actually gets sent. That second part is non-negotiable. I wrote about what happens when AI agents send email without oversight and it’s not pretty.
What the demo shows
TextForge is an MCP server (although we also have a rich HTTP API too) that plugs into Claude Code. In the video, I run a single command from a Claude Code skill, /sales-inbox, and the agent goes to work.
I wrote a fully-featured Pipedrive CLI to help automate this. It works on all platforms. All it requires is a Pipedrive API key. This makes it really easy to get Claude Code and other LLM harnesses to automate sales alongside TextForge.
It starts by connecting to my Pipedrive CRM and scanning my active deals (0:36). Deal stage, last contact date, deal notes, and so on. Claude figures out who’s due for a message and what to say based on what the CRM says. Then Claude uses TextForge’s search_messages / etc MCP methods to see if there’s been any updates that haven’t been accounted for in Pipedrive yet.
If it looks like we are due for a follow up, Claude drafts personalized follow-ups for each one in TextForge. Not templates with {first_name} swapped in. Actual emails that reference the specific deal, attach the right document, and read like something I’d write.
The approval step is the whole point (1:20): once the draft is ready, I get a Slack notification. I can skim it from my phone and approve in seconds, or open the TextForge dashboard to read the full email and make edits (or I can ask Claude to make them, which it can also do.) In the demo I tweak a line and hit approve (1:39). Maybe 10 seconds per email.
The approved email sends from my actual Gmail account, with my name and signature (2:10). The recipient has no idea an AI was involved. And when they reply, TextForge picks it up via webhooks and pings me in Slack. I can have the agent draft a response, or jump in myself.
The results
Since I started using this workflow, I’ve cut those 8-10 hours per week down to about 30 minutes of review time per day. The emails are honestly better than what I was writing manually (short, concise, and direct.) Claude is more consistent about pulling in deal-specific details, and it never forgets to attach the right document.
The part that surprised me: revenue from my sales pipeline is up about 20%. Not because the emails are magic, but because follow-ups actually happen consistently now. Deals that used to go cold because I didn’t get around to the email on a busy week now get a follow-up within 24 hours. Consistency beats perfection.
Try it yourself
The sales automation workflow is one of the most popular TextForge use cases, but the same pattern works for any email workflow where you want AI doing the drafting and a human doing the deciding.
If you want to set this up for your own pipeline, the quickstart guide walks you through connecting your email, CRM, and notification channels. Setup takes about 10 minutes.
TextForge has a 7-day free trial. Connect your tools, run a few drafts, and see if the approval flow works for how you actually operate. That’s the real test: not whether the AI can write emails, but whether the review step is fast enough that you’ll actually use it every day.
It is for me. That’s why I built it. Now you can use it too.