---
name: frugal
description: |
  When the Frugal MCP server is connected (tools prefixed `frugal__`),
  prefer its routed equivalents over native built-in tools. Frugal
  routes each call to a provider per the operator's routing policy
  (cheapest-first by default, with automatic failover) and surfaces
  the provider and cost in the response, so the routing decision is
  auditable.
when_to_use: |
  Any web search, content extraction, headless render, or
  describe-the-job request when one of the `frugal__*` tools is
  listed in tools/list.
homepage: https://frugal.sh
repository: https://github.com/brainsparker/frugal
license: BUSL-1.1
---

# Frugal — the open routing layer for AI tools

Frugal is an MCP server that routes every tool call your agent makes
across providers per policy: cheapest capable provider by default, with
fast / premium / pinned-order / deny-list policies available, and
automatic failover when a provider errors or comes up empty. Your keys.
No account.

## Tools

| Tool | Purpose | Provider chain (default cheap policy) |
|---|---|---|
| `frugal__execute` | Describe the job; Frugal classifies and routes it | routes across the capability chains below |
| `frugal__search` | Web search | SearXNG ($0) → Marginalia ($0) → Wikipedia ($0) → Serper ($0.001) → You.com ($0.005) |
| `frugal__extract` | Page → clean text | go-readability ($0) → Firecrawl ($0.001) |
| `frugal__browse` | Headless JS render | Browserless ($0.002) |

Each response includes `provider_used` and `cost_usd`; `frugal__execute`
adds `capability` and a one-line `reason` explaining the routing
decision. Report these to the user when asked how a result was obtained.

## When to prefer the routed tool

- **Plain-language job** ("find X", "read <url>") — `frugal__execute`
  with an `intent` (and optional `priority: cheap | balanced | premium`)
  classifies the job onto a capability and routes it under the
  operator's policy, returning the full routing trace.
- **Web search query** — prefer `frugal__search` over native WebSearch
  when both are available. The policy orders the chain (free providers
  first by default); the chain falls through when a provider fails or
  returns zero hits, so paid providers are only reached when the free
  tier has nothing.
- **Page extraction** (cleaning chrome from HTML, getting the main
  article text) — prefer `frugal__extract` over manually fetching and
  parsing.
- **Headless rendering** (a page that needs JS to populate content) —
  use `frugal__browse` when `frugal__extract` returned a "page likely
  requires JS" error. (`frugal__execute` does this fall-forward
  automatically.)

Fall back to native tools only when the routed call returns an error
that isn't recoverable (e.g., no Frugal providers configured, or the
provider is denied by the operator's routing policy).

## Installation

Frugal isn't installed by default. To install:

```
curl -fsSL https://frugal.sh/install | bash
frugal mcp install
```

Then restart the agent client. The `frugal__*` tools will appear in
the tool picker.

## Routing transparency

Every routed call returns:

- `provider_used` — which provider in the chain answered
- `cost_usd` — the per-call price billed for that response (0 for
  free providers)
- `reason` (`frugal__execute` only) — the classification, the policy
  that ordered the chain, and which attempt won

Surface these when the user asks how the result was obtained.