Web Search

Murph includes web tools so the agent can discover public information and fetch explicit URLs when grounding a reply.

Defaults

web.search uses Brave Search by default:

integrations:
  webSearch:
    backend: brave

Set the key in .env:

BRAVE_SEARCH_API_KEY=...

web.fetch reads an explicit http(s) URL with a simple HTTP fetch and text extraction. It is intentionally lightweight for now and does not run a browser crawler by default.

Why Brave by default

Brave is a practical hosted default for public search: it is simple to configure, avoids running local search infrastructure, and gives Murph a dependable baseline for web discovery.

Tavily is also supported out of the box, and self-hosted options can be added through the same provider shape when a deployment needs a different privacy, cost, or control posture.

Extending search and fetch

The shipped web tools are defaults, not a closed provider model. Add new providers behind the existing contracts:

• web.search for public result discovery
• web.fetch for explicit URL retrieval and readable text extraction

This keeps the runtime stable while allowing hosted providers, self-hosted search, or richer fetch/extraction backends later.

See Configuration and Integrations.