Crush
Crush's built-in web_search tool queries DuckDuckGo. Add SpecStream with one mcp add command in your crushrc, and Crush can search the live web, read whole pages, and build lists with a source on every value.
What you need
- Crush v0.100.0 or newer.
- A SpecStream account. New accounts get $10 of credit, no card.
- 1
1. Add SpecStream to your crushrc
Open
~/.config/crush/crushrc, or%USERPROFILE%\.config\crush\crushrcon Windows, and add these lines. To add SpecStream to one project only, put them in./crushrcor./.crushrcat the project's root instead.mcp addadds the server, or updates the one namedspecstreamif it exists.--oauth trueturns on Crush's OAuth 2.1 sign-in, so there is no key to paste.crushrc - 2
2. Sign in to SpecStream
Start Crush. It opens a dialog titled Authenticate with specstream. Press enter to open your browser, sign in to SpecStream or create an account, and approve the access Crush asks for.
Crush shows Authentication successful! and keeps the token for the next session.
- 3
3. Ask Crush
Ask for something current and name SpecStream. Crush calls
search_webfor a quick answer andread_pagesfor the full text of the pages it found:In Crush - 4
4. Or connect with an API key
Use a key on a machine with no browser. In the workspace, open Agents & API, choose API keys, then Create API key, and set a total limit. Export it as
SPECSTREAM_API_KEYwhere Crush starts, and send it as a header in place of--oauth.Crush drops a header whose value is empty, so an unset variable sends no key at all.
crushrc with a key
What Crush gets
SpecStream's server at https://api.specstream.ai/mcp adds these tools to Crush.
search_webanswers a question from the live web, with up to 10 results for $5 per 1,000 searches. A search that returns nothing costs nothing.read_pagesreturns the main text of up to 20 pages or documents (PDF, Word, Excel) in one call, at $0.80 per 1,000 pages. A page that gives no text is not charged.- Deep Search builds a list from a request in plain words.
compilereads the request back,quote_capacity_workprices it for free, andauthorize_capacity_workstarts it only after you approve the maximum.get_findandlist_find_rowsread the rows, each with its source.
Crush's own web search and SpecStream
Crush ships a web_search tool that queries DuckDuckGo, and no setting points it at another search engine. SpecStream runs beside it as an MCP server, so both stay available. Name SpecStream in your request when you want its results.
SpecStream's Search API scores 97.9 F-score on all 1,000 SimpleQA Verified questions and 64.0 item F1 on all 200 WideSearch tasks. The method is on benchmarks.
Approvals in Crush
Crush asks before it runs a tool you haven't allowed, for example with permissions allow in crushrc. Searches and page reads are charged per call with no quote. Deep Search adds its own gate: nothing paid starts until you approve the price SpecStream quotes.
Cost
Connecting SpecStream to Crush is free, and new accounts start with $10 of credit. A search of up to 10 results costs $5 per 1,000, and a page read costs $0.80 per 1,000. Deep Search is priced per delivered row and quoted before it runs, and a run can't cost more than the maximum you approved. See pricing and how billing works.
Remove
Delete the mcp add specstream lines from crushrc and restart Crush. To end its access, open Agents & API in your SpecStream workspace and revoke Crush under OAuth applications, or revoke the key under API keys. Questions: [email protected].
More
SpecStream for Crush covers what it does in Crush on one page. Every other tool has its guide in Connect your tools. You see the most a request can cost before it starts, and the matches are paid from your balance. Billing has the details. For help, email [email protected].