# fetchira

> Fetchira is an open source (Apache 2.0) MCP server and CLI that puts 24 web search, scrape, research and image providers behind six tools: search, read, deep_research, browser, create_image and usage. It sends each call to an account with quota left and fails over on errors. It runs on your machine or on a server you host; fetchira.dev is its website and does not run a public MCP endpoint.

Install with `npx -y fetchira` (macOS and Linux, needs Node.js), `brew install ImmuneFOMO/tap/fetchira`, or `curl -fsSL https://raw.githubusercontent.com/ImmuneFOMO/fetchira/main/install.sh | sh`. Then `fetchira add <provider>` and `fetchira install` to register it in Claude Code, Codex, Cursor, Gemini CLI, OpenCode, Zed or Windsurf. Every fetchira.dev page also answers `Accept: text/markdown`, or add .md to its path (/index.md for the home page).

Providers: search (serper, tavily, exa, keenable, tinyfish, langsearch, parallel, you, linkup, nimble, octen, serpapi, openalex, gdelt); read and scrape (firecrawl, jina, scrapingant, zenrows); browser (steel, hyperbrowser, browserless); web sessions (chatgpt_web, gemini_web, grok_web).

## Docs

- [Home](https://fetchira.dev/index.md): what fetchira does, install, the six tools, providers and FAQ
- [Overview](https://fetchira.dev/docs.md): install, first search, and which guide covers what
- [Setup](https://fetchira.dev/docs/setup.md): install, connect an agent, choose a skill
- [CLI](https://fetchira.dev/docs/cli.md): search, research and images from a shell
- [Configuration](https://fetchira.dev/docs/configuration.md): accounts, quotas, proxies, troubleshooting
- [Hosted deployment](https://fetchira.dev/docs/hosted.md): share one private server through CLI or MCP
- [Full text](https://fetchira.dev/llms-full.txt): every page on this site, docs included, in one file

The docs pages are synced from https://github.com/ImmuneFOMO/fetchira at build time; the repository is the source of truth.

## Agent skills

- [fetchira](https://fetchira.dev/.well-known/agent-skills/fetchira/SKILL.md): Use Fetchira for current web search, URL reading, deep research, JavaScript page reading, image generation, and provider quota routing. Prefer its listed MCP tools and use the CLI only when MCP is unavailable.
- [fetchira-cli](https://fetchira.dev/.well-known/agent-skills/fetchira-cli/SKILL.md): Use the Fetchira CLI for current web search, URL reading, deep research, JavaScript page reading, image generation, and provider usage through a shell.
- [fetchira-mcp](https://fetchira.dev/.well-known/agent-skills/fetchira-mcp/SKILL.md): Use the configured Fetchira MCP tools for current web search, URL reading, deep research, JavaScript page reading, image generation, and provider usage. Use only when Fetchira MCP is exposed by the host.
- [Skills index](https://fetchira.dev/.well-known/agent-skills/index.json): Agent Skills Discovery v0.2.0, archives with SHA-256 digests

## API

- [Providers](https://fetchira.dev/api/providers.json): JSON list of all providers with capabilities, auth type and free allowance
- [OpenAPI](https://fetchira.dev/api/openapi.json) and [API docs](https://fetchira.dev/api/docs.md)

## Blog

- [Why fetchira puts 24 providers behind six tools](https://fetchira.dev/blog/why-fetchira.md): One MCP server per search provider costs about 95k tokens of tool definitions on every request. Fetchira does the same work with six tools and 2.7k tokens.

## Optional

- [Source code](https://github.com/ImmuneFOMO/fetchira)
- [Changelog](https://fetchira.dev/changelog.md)
- [Security policy](https://fetchira.dev/security.md): report vulnerabilities privately
- [License](https://fetchira.dev/license.md): Apache License 2.0
- [Privacy](https://fetchira.dev/privacy.md): this site sets no cookies and runs no analytics