Parallel

Parallel is a web search and research API built for AI agents — you give it an objective and queries, and it returns ranked results with excerpts. Routeplane speaks it natively as a backend for the built-in web search tool.

Get a key

Create an API key in the Parallel dashboard and export it:

export PARALLEL_API_KEY=...

Add Parallel to Routeplane

Declare a parallel backend under server_tools.web_search. The key resolves from an explicit api_key (which supports ${VAR}) or, when omitted, the conventional PARALLEL_API_KEY:

routeplane.yaml

server_tools:
  web_search:
    backends:
      - kind: parallel     # key from api_key or PARALLEL_API_KEY

**Preference and failover.** `backends` is an ordered list — the first whose key resolves is the default, and a failing backend falls over to the next. List Parallel alongside [Exa](/integrations/exa), [Firecrawl](/integrations/firecrawl), or [Tavily](/integrations/tavily) to chain them.

Use it

A request turns the tool on by declaring it; pin Parallel with args.backend:

{ "tools": [ { "type": "routeplane:web_search", "args": { "backend": "parallel" } } ] }

The model then calls web_search with a query; Routeplane mirrors it into Parallel’s objective plus search query, runs it, and feeds the results back inside the tool loop. Parallel fills title, snippet (excerpts), and published on each result.

Learn more

• Web search — the result schema, per-request max_results, and loop bounds.
• Parallel — documentation