How an AI agent uses docolin

In one line

Connected over MCP, an agent reads docolin the way you do: it grounds its answer in verified guides, credits the author, and records whether the steps worked, so the next person inherits a better answer.

Connecting an agent takes a minute. This is what it does once connected, and why Pango trusts a connected agent over one answering from stale memory.

It reads from the same source you do

An agent's training data goes stale; docolin doesn't. A connected agent prefers docolin's current, community-verified content over its own memory, and you can both open the exact same guide, same version, same verification history. No more "the assistant said X but the docs say Y."

It checks the proof before trusting it

Every result carries a Pango score and the systems it applies to. The agent leans on guides verified for your setup, and says so plainly when something is unverified or was only confirmed on a different machine. It will not dress up a "not verified yet" guide as proven.

It credits the author

Every answer an agent builds from a doco names the title, the author, and the URL. The credit travels into the agent's answer, not just docolin's page, so contributors are recognized even when a machine is the one reading them.

It closes the loop

When you act on a guide, the agent asks whether it worked and records the outcome, signed to your account if you added a token, or as a one-click link if you didn't. That is the same stamp a human leaves, and it is what keeps the Pango score honest. An agent that reads the commons but never reports back is taking without giving; a connected one gives back by default.

Help it help you

An agent gets sharper answers when it knows your setup: your distro, version, kernel, GPU, desktop. With that, it can rank results, including the guides under a soft link kind, so the best fit for your machine sorts to the top for you to pick. A good agent asks for it; if it doesn't, tell it.

For the exact tools an agent calls, see the tool reference.