Coding agents are good at writing code. They are not good at showing you things.
When an agent finishes a task it typically has nowhere to put the result. It might paste a wall of text into the chat, write a file somewhere you have to go find, or just tell you it's done and hope you believe it. None of these are satisfying. None of them let you review what happened, ask a follow-up question, or hand the work off to someone else.
pail is a hosted set of MCP-native primitives for exactly those moments.
It is a bucket your agent can put things in. A file, a diagram, a diff awaiting your approval, a live progress update, a question blocking until you answer it. Each one becomes a URL. Each URL opens in a browser as a clean, rendered page.
Who it's for
pail is for developers who run coding agents — Claude, Codex, Cursor, Cline, or anything else that speaks MCP. If you've ever wanted your agent to show you a rendered diff instead of a raw patch, or ask you a clarifying question and wait for an answer, or post a status page you can check mid-run, pail is for you.
It is not for end users. There is no GUI for creating content by hand. The interface is an MCP server and a REST API. Humans interact with pail by opening URLs that agents give them.
How it works
Add pail to your agent with a single command:
claude mcp add pail \
--transport http \
--header "Authorization: Bearer $PAIL_API_KEY" \
https://pail.thalos.ai/mcp
Now your agent has ten new tools. Each tool takes content in and returns a URL. You open the URL. That's it.
The free tier is 30 actions per day, 5 GB of storage, 30-day TTL, and every tool except preview — no credit card required. Upgrade to Pro ($9/mo) for 200 actions/day, 30 GB storage, 90-day TTL, and preview.