Searching the guide
Search the guide from your terminal with seedcord ask, using a question, a pasted error, or an error code, and get links to the sections that match.
seedcord ask searches the guide from your terminal. Use it to look up an error you just hit, or to check whether seedcord already ships something before you write it yourself. It's also great for agents as they're able to query the guide programmatically on the spot.
It runs the search behind the Related results in the guide's search dialog. That search matches by meaning, so your question doesn't have to use the guide's words.
pnpm exec seedcord ask how do i stop someone spamming a commandThe answer lists each matching page with its matching sections, and a link to each section. You can run it outside a project too, before your bot exists.
Quote the question if it contains --verbose or --help, which the CLI reads as flags, or a character that your shell treats as special, like ? or [ in zsh. Other words that look like flags stay in the question.
A stable release of the CLI searches seedcord.org/guide. A prerelease searches next.seedcord.org/guide, because that site has the pages for code that isn't in a stable release yet.
Pasting an error
Paste the whole error as the question. seedcord ask removes the stack frames, then searches the first 500 characters of what remains. If the command shortens your text, the answer says so.
A seedcord error starts with its code in brackets, like [3116]. The guide's pages use the error's name, so the command looks up the name for you:
| you run | what it searches for |
|---|---|
seedcord ask "[3116]" or seedcord ask 3116 | CliDependencyMissing |
seedcord ask error 3116 | error 3116 CliDependencyMissing |
The answer starts with 3116 is CliDependencyMissing. Inside a longer question, a four-digit number stays as you wrote it, since it could be a port, year, or something else entirely. If that number is also an error code, the answer adds the command that looks up the code.
Running it from a script or an agent
With no terminal attached, or with CI set, seedcord ask prints only the answer. Each page in it links to its Markdown copy and lists the matching headings, so that a coding agent can read the page without its HTML. If nothing matches, the answer links llms.txt, the list of every page in the guide. If the search fails, the command exits with code 1.
A project from create-seedcord tells coding agents to use this command in its AGENTS.md.
When it fails
| what happened | the error |
|---|---|
| your question is under 2 characters, or doesn't contain a letter or a number | CliAskBadQuestion |
| the guide search got too many requests or hit its daily limit, and the message says when to try again | CliAskHasToWait |
| the request failed, or took longer than 10 seconds | CliAskFailed |
CliAskHasToWait and CliAskFailed print the guide's address, so you can open the guide and read it yourself.