@kivlor/pi-websearch
Exa-backed web search and content fetch for pi — raw TypeScript, no build step
Package details
Install @kivlor/pi-websearch from npm and Pi will load the resources declared by the package manifest.
$ pi install npm:@kivlor/pi-websearch- Package
@kivlor/pi-websearch- Version
0.1.3- Published
- Oct 7, 2026
- Downloads
- 598/mo · 136/wk
- Author
- kivlor
- License
- MIT
- Types
- extension
- Size
- 20.4 KB
- Dependencies
- 0 dependencies · 2 peers
Pi manifest JSON
{
"extensions": [
"./index.ts"
]
}Security note
Pi packages can execute code and influence agent behavior. Review the source before installing third-party packages.
README
@kivlor/pi-websearch
Web search for the Pi coding agent, powered by Exa: search the web and fetch full text content for known URLs.
Install
pi install npm:@kivlor/pi-websearch
--local/-l writes the declaration to the project's .pi/settings.json
instead of the personal one. Manage installed extensions with pi config.
Tools
web_search— search the web with Exa. Singlequeryor up to 4 variedqueriesrun concurrently; optionalnumResults,includeContent,recencyFilter(day/week/month/year),domainFilter(-domainexcludes).web_fetch— fetch full text content for known URLs via Exa's/contentsendpoint (up to 10 per call).
Usage
Ask pi: "search the web for recent news about X" or "fetch the full content of this URL", and the model calls the matching tool.
Commands
/websearch— show Exa configuration status.
Configuration
Resolution order:
exaApiKeyin~/.pi/agent/web-search.jsonEXA_API_KEYenvironment variable
With a key, searches hit the Exa REST API (api.exa.ai, override with
exaBaseUrl in config or EXA_BASE_URL). Without a key, they fall back to
the free Exa MCP endpoint (mcp.exa.ai), which is rate-limited (429s).