The inbox where your agents
hand you decisions.
Pileless lets your autonomous agents pause only for the calls a human should make — it collects those questions in one inbox, and your agent resumes the moment you decide. Built for agents that run unattended; chat already handles the case where you're sitting right there.
0037_add_index.sql · CI green, rollback verified. Agent suggestion: approve (additive, no destructive ops).Chat solved the synchronous case. Pileless owns the async one.
In a live chat you're already there — you can see the image and answer the question inline. Pileless earns its keep precisely when no human is in the loop: an agent grinds through work you never watched, then hands you a plate of decisions to clear later.
Works while you sleep
Agents run long, autonomous jobs and pile up the decisions. You wake to a sorted queue, not a mess.
Decide in seconds
Approve, reject, pick one, rank, edit — in place. Blocking items on top, fastest-to-slowest below. Your order, drag-to-customize.
Machine-readable back
Every decision returns to your agent as structured data so the workflow resumes itself — no copy-paste, no babysitting.
What your agents stop and ask you
The handful of calls you don't want an agent making alone — collected, not scattered across logs and Slack pings.
🚀 Ship it?
"Deploy v2.14 to prod — CI's green, here's the diff." Approve, request changes, or reject.
🔀 Merge this PR?
"Refactored the Stripe client, 4 files, tests pass." Sign off or send back with a comment.
✉️ Send this reply?
"Drafted a response to the angry customer." Approve, edit in place, or rewrite.
🧩 Pick the option
"Three auth libraries — Lucia, Auth.js, WorkOS. Which do I build on?" Choose one.
💸 Approve the spend?
"Vendor invoice, $18,450, PO matches." Yes/no before money moves.
🛠️ Apply this config change?
"New env var + index, here's the plan." Approve before it touches infra.
Bring your own brain. We run none.
Pileless doesn't run any AI on your content. Your agent — Claude Code, Codex, any model on your key — drives it over MCP or the API. We route, render, validate, and enforce policy, and we never use your content for training, mining, or sale.
Plug in any agent
Claude Code and Codex wire up in a minute via MCP. Your agent calls pile.create and waits for the human's call.
Your files, your cloud
Bring-Your-Own-Cloud keeps documents in your storage bucket — file objects never sit on Pileless infrastructure.
Least-privilege by default
Your agent sees only the fields you clear it for; sensitive content is redacted server-side. Every access is audited.
Connect your agent in a minute
Bring your own cloud
Keep file objects in your own Cloudflare R2 bucket — your storage, your control.
Stop babysitting your agents.
Let them work, and let Pileless hand you only what needs a human.
Get started →