Adding an AI copilot to your app means picking a frontend SDK, an agent framework, and a vector database. Then wiring auth, streaming, evals, guardrails, chunking pipelines between all of them. Or you can just use Pillar. That's what we've built.

Setting up a single trigger in Zendesk takes 30 clicks With Pillar it takes one sentence @jjmaxwell4 built an open source copilot you build into your app. Users talk to it in natural language and it drives the app for them The problem: products can do a lot but users don't

We're going to see more and more companies launch copilot agents like this in 2026. Stripe, Airbnb are next.

a user asked our copilot to help them cancel their subscription. it worked flawlessly. 5 stars. not sure how to feel about this.

Read more here:

That is how valuable "replace dashboard clicks with code" became for infrastructure. Pillar is doing that for product UX. Actions users describe in one sentence should be defined in code, reviewed in PRs, and run with the user's session.

In the 2010s, Terraform turned infrastructure changes into code. By 2021, the company behind it, HashiCorp, was worth about $14B in the public market.

Announcing AgentMail's $6M Seed led by @generalcatalyst No pressure, right?

Thrilled to share a major milestone at @trypillar_ai . Yesterday: we didn't have billing. Today: $19 in recurring revenue. If you extrapolate from the last 4 minutes — which is absolutely a valid methodology used by top analysts — we're running at a $2.5M ARR rate. That's a

Your users already know what they want. They just don't know which menu it's buried in. We're building @trypillar_ai so they can say it instead of clicking through it.

Same thing if Zendesk used Pillar: "Route all billing, invoice, and payment tickets to Finance with high priority" 14 words. Done.

Scroll down to Actions. Add one. Pick Priority → High. Add another. Pick Group → Finance. Save. 30 clicks for one trigger. Most Zendesk instances have 20-50 of these.

Admin Center → Objects and rules → Triggers → Add trigger → Name it → Add condition → pick Subject → pick Contains → type "billing" → Add condition → pick Subject → pick Contains → type "invoice" → Add condition...

I counted every click it takes to set up one ticket routing rule in Zendesk. 30 clicks! 🧵

When the user clicks "Reveal," the SDK fetches the key directly from your server. One redemption, then the reference burns. Try it at

Your backend creates the key, encrypts it, and returns a short-lived reference token. The agent only sees { secret_ref: "ref_a8f3..." }

In most architectures, tool outputs pass through a reasoning server. The key ends up in logs, conversation history, and third-party infrastructure you don't control.

Your AI agents need to hand API keys to users. Here's how we handle that without the key ever touching the AI pipeline.

The bar for cold outbound in 2026: a personalized Loom walking through a working demo built for that clients specific use case. Not a template. Not an AI-written email. A real thing build for a specific client. It's proof of work. Spam ruined everything below that line. The

🚀 @trypillar_ai launched! Your App's Copilot "Pillar is an open source copilot that turns user and agent requests into completed actions, right inside your app." 🌐 https://t.co/jRGP6B8Nlf Congrats @jjmaxwell4 & Mark Wai!!