Skip to main content
Glama
README.md
![Fetchira](docs/fetchira.webp)

# Fetchira

Web search, page reading, deep research and image generation through one router. Use the
CLI or connect your agent through MCP. Bring an API key or a supported web session;
Fetchira routes requests across your configured accounts and providers.

![One tool list for every provider](docs/providers.png)

Fetchira covers 24 providers, 21 through APIs and 3 through browser sessions, behind six MCP
tools. Connected one by one, the 17 providers that publish an official MCP server add about
95k tokens of tool definitions to every request, a median of 3.7k per server; the whole
Fetchira tool list is about 2.7k. One router picks the account with quota left, fails over on
errors and keeps every provider's limit in one place. Provider names and logos belong to their
respective owners.

## Start locally

Install and open the dashboard with one command (macOS or Linux, needs Node.js):

```sh
npx -y fetchira
```

Or install on macOS with Homebrew:

```sh
brew install ImmuneFOMO/tap/fetchira
```

On Linux or macOS without Homebrew:

```sh
curl -fsSL https://raw.githubusercontent.com/ImmuneFOMO/fetchira/main/install.sh | sh
```

See [installation](docs/setup.md#install) for Linux arm64, building from source and PATH setup.

Add a provider and make a request:

```sh
fetchira add serper --key 'YOUR_SERPER_KEY'
fetchira search "latest Rust release"
fetchira read "https://example.com"
```

Get a Serper key at [serper.dev](https://serper.dev), or run `fetchira providers` to choose
another provider. Each provider has its own plan and quota; paid features can consume a
paid balance. For browser login instead of an API key, run `fetchira add gemini_web`.

MCP is optional. Run `fetchira setup` to choose local accounts or an existing hosted server.
For an agent, `fetchira install` offers CLI only, MCP, or MCP with CLI fallback, then installs
the matching skill in your selected agents. Restart the agent after installation.
`fetchira ui` opens the local dashboard.

![Local dashboard](docs/dashboard.png)

## Choose a guide

| I want to… | Guide |
|---|---|
| Install, connect an agent, or choose a skill | [Setup](docs/setup.md) |
| Search, research, generate images, or continue a session from a shell | [CLI](docs/cli.md) |
| Manage accounts, quotas, proxies, or troubleshoot a failure | [Configuration](docs/configuration.md) |
| Share one private server through CLI or MCP | [Hosted deployment](docs/hosted.md) |
| Build, test, contribute, or release | [Contributing](CONTRIBUTING.md) · [Releasing](docs/releasing.md) |

## Hosted

To use an existing server, get its URL and API key from the administrator. Run
`fetchira setup` and choose **Connect to a server** to verify access before saving.
For scripted setup, save the connection and check it:

```sh
fetchira remote set https://fetchira.example.com/mcp --key 'fk_live_...'
fetchira remote check
```

CLI and MCP both use the configured server; generated images and PDFs are saved on your
computer. Laptop file attachments require local mode through either interface.
See [hosted deployment](docs/hosted.md) to run your own server.

[Apache-2.0 license](LICENSE) · [Report a security issue](SECURITY.md)

TDQS

A3.9/5.0

Scored across 6 tools

Disambiguation4/5

Most tools target clearly different actions (search, deep_research, create_image, usage), but browser and read overlap heavily since both load a URL and return markdown. The read description's note that it auto-escalates to browser helps, but an agent could still reasonably pick either for a simple fetch.

Naming Consistency3/5

All names are lowercase snake_case and readable, but conventions are mixed: bare verbs (read, search), bare nouns (browser, usage), a verb_noun (create_image), and a compound (deep_research). No single predictable pattern guides naming.

Tool Count5/5

Six tools is well within a healthy range and each earns its place: fetching, searching, deep research, image generation, and quota inspection. No obvious filler or missing companion tool.

Completeness4/5

The surface covers the core domain well: URL fetch via read/browser, search, multi-source research, image generation/editing, and quota introspection. Minor gaps include no explicit batch/multi-URL crawl tool (only available via provider-specific modes) and no dedicated session or result management beyond opaque tokens.

Maintenance

ActivityActive
ResponsivenessNo recent community issues