An agent that answers with a table, not a paragraph

Any MCP client and Pando, for structured results you can sort.

Ask an assistant to compare six things and you get six paragraphs. The comparison is in there, and reading it means holding six shapes in your head at once to find the one column you actually care about.

Ask it to write the result as a bullet with the table layout: the children become rows, their children become cells, and the first row names the columns.

Set it up

Point the client at https://pando.ink/mcp, which speaks streamable HTTP.
Authenticate with OAuth 2.1 and dynamic client registration, or paste a key from Agents and API keys as a Bearer token.

The shape, in one picture

Vendors compared          layout: table
  Name | Hosted | Export     <- the first row names the columns
  Alpha | yes    | OPML
  Beta  | no     | none

Three levels: the bullet carries the layout, each child is a row, and each child of a row is a cell. That is also why a table is three calls rather than one, and the tool description says so before the agent tries.

What you get that a chat table does not give you

It is still an outline. Every cell is a bullet, so a cell can carry a note, a link or children of its own without breaking the grid.
You can edit it. A wrong cell is fixed by typing in it, not by asking for the whole table again.
It exports as a table. Read the branch as Markdown and it comes out as a GitHub-flavoured table, by the same writer the export door uses.
Set the layout back to bullet and nothing is lost. The rows are still there as an ordinary branch.

When to ask for a table and when not to

Ask for one when the answer has the same fields repeated: vendors, files, test failures, candidates, prices. Do not ask for one when the answer is an argument. A table of reasons is a way of hiding that the reasons do not line up, and both of you will read it as though they do.

Deeper

Connect an agent, about two minutes · The twelve tools · Which note apps an agent can reach · Who runs this