v0.1.0

seek v0.1.0

seek is a web tool for CLI coding agents.

It provides a single interface over multiple search, scraping, and crawling providers, with automatic failover so agents can continue working even when a provider is unavailable or rate-limited.

Highlights

• Automatic provider failover — search and scrape requests transparently fall back to the next configured provider.

• Seven providers, one interface — supports Firecrawl, Tavily, Spider.cloud, WebCrawlerAPI, Lightpanda, Brave Search, and Exa.

• Agent-first design — includes:

  • MCP server (seek mcp)

  • HTTP JSON API (seek serve)

  • Claude Code skill integration

• Token-efficient workflows — CSV-first output and snippet-before-scrape patterns help reduce agent costs.

Commands

Command | Description -- | -- seek search | Search the web seek scrape | Extract a page as markdown, HTML, or JSON seek crawl | Crawl a website seek serve | Run the HTTP JSON API seek mcp | Start an MCP server over stdio seek config init | Configure providers and API keys seek version | Show version information and check for updates

Installation

Install Script (Linux/macOS)

curl -fsSL https://raw.githubusercontent.com/Rishang/seek/main/install.sh | sh

mise

mise use -g ubi:Rishang/seek

ubi

ubi --project Rishang/seek --in ~/.local/bin

After installation:

seek config init

Then connect your agent using the included skill or register seek mcp.

Notes

• crawl currently uses Firecrawl directly and does not support failover.

• seek serve is unauthenticated unless --token or SEEK_SERVE_TOKEN is configured.

• Provider API keys are stored separately from configuration and can be overridden with environment variables.

New Contributors

• @Rishang made their first contribution in #1

Full Changelog: https://github.com/Rishang/seek/commits/v0.1.0