> SpecStream is the live web data engine for teams and agents. This is the Markdown copy of https://specstream.ai/web-search-mcp. Every page, with what SpecStream does and costs: https://specstream.ai/llms.txt

# MCP web search servers, compared | SpecStream

Compared on Nov 8, 2026.

# Web search over MCP

SpecStream's MCP server gives any agent the most accurate web search API in the world. SpecStream leads WideSearch (60.9, on the low-cost model; the best other API scores 57.6 even on the larger model) and SimpleQA Verified (99). See the [benchmarks](https://specstream.ai/benchmarks?utm_source=alternatives&utm_campaign=web-search-mcp). Add `https://api.specstream.ai/mcp` to Claude Code, Cursor, or any MCP client, and the agent can call `search_web`, read pages with `read_pages`, and build structured rows with Deep Search, on one sign-in and one balance. SpecStream is the live web data engine for teams and agents.

[Get an API key](https://specstream.ai/signup?src=alternatives-web-search-mcp&utm_source=alternatives&utm_campaign=web-search-mcp&returnTo=%2Fworkspace%2Fagents%3Ftab%3Dkeys)

New accounts get $10 of credit for their first 30 days, no card.

## Six MCP servers for web search, side by side

|  | SpecStream | Tavily | Exa | Perplexity | Parallel | Brave |
| --- | --- | --- | --- | --- | --- | --- |
| WideSearch, low-cost model | 60.9 | 47.9 | 53 | 47 | 45.5 (Fast), 44 (Turbo) | Not in the published scores |
| SimpleQA Verified, larger model | 99 | 92 | 91 | 95 | 97 (Advanced, Basic) | Not in the published scores |
| Remote endpoint | api.specstream.ai/mcp | mcp.tavily.com/mcp/ | mcp.exa.ai/mcp | api.perplexity.ai/mcp | search.parallel.ai/mcp (labeled legacy) | None stated; a local npx server |
| Claude Code install | One `claude mcp add` line | One `claude mcp add` line | A plugin install | One `claude mcp add` line | One `claude mcp add` line | Not stated |
| Search tool | `search_web` | `tavily_search` | `web_search_exa` | `perplexity_search` | `web_search` | `brave_web_search` |
| Page reading tool | `read_pages`, $0.80 per 1,000 pages; a page with no text is not charged | `tavily_extract` | `web_fetch_exa` | Not stated | `web_fetch` | Not stated |
| Research tool | Deep Search: structured rows at 5¢, 15¢ or $1 a delivered row | `tavily_research` | `agent_run` | `perplexity_research` | `web_research` in the main Parallel MCP | Not stated |
| Price per 1,000 searches | $5 | $8 (basic), $16 (advanced) | $7 (auto, fast), $4 (lite, instant) | $5, $1 (Fast Search) | $5 (basic, advanced), $1 (turbo, fast) | $5 |
| Spending cap before a run | A quote with a maximum you approve | Not stated | Not stated | Not stated | Not stated | Not stated |

## Why agents search over SpecStream's MCP server

### The search tool decides the answer

An agent trusts what its search tool returns. With SpecStream, an agent on the low-cost model scored 60.9 on WideSearch, above every other API in the published scores on either model, and 99 on SimpleQA Verified on the larger model. `search_web` runs on the same Search API, at the same $5 per 1,000 searches.

### One server for the whole engine

`search_web` answers a question with a value, the page it came from, and the time it was read. `read_pages` reads pages as markdown at $0.80 per 1,000, and a page with no text is not charged. For a dataset, the agent calls `compile` to read a request in plain words, `quote_capacity_work` to price it, and `authorize_capacity_work` to run it. Rows cost 5¢, 15¢ or $1 each, plus 5¢ for an email and 15¢ for a phone, and a run that finds nothing costs nothing.

### Spending the agent cannot exceed

Searches and page reads run at their list price with no approval step. Deep Search runs only after you approve the maximum in its quote, passed as `confirm_max_cents`, and the run stays under it. Saved requests rerun from hourly to yearly, with changed rows at half rate and unchanged rows at 1¢, and results go to Slack, email, a sheet, a CRM, or a signed webhook.

## How SpecStream scores

The same numbers as the benchmarks page, from the official tasks and scorers.

ResultsWideSearchCostSimpleQA Verified

|  | SpecStream | ![](https://specstream.ai/images/benchmarks/parallel.svg)Parallel | ![](https://specstream.ai/images/benchmarks/exa.svg)Exa | ![](https://specstream.ai/images/benchmarks/tavily.svg)Tavily | ![](https://specstream.ai/images/benchmarks/perplexity.svg)Perplexity |
| --- | --- | --- | --- | --- | --- |
| **Web research**WideSearch, GPT-5.6 Luna | 60.9 | 45.5 | 53.0 | 47.9 | 47.0 |
| **Factual accuracy**SimpleQA Verified, GPT-5.6 Luna | 96 | 94 | 91 | 94 | 94 |
| **Factual accuracy**SimpleQA Verified, GPT-5.6 Sol | 99 | 97 | 91 | 92 | 95 |

## How to add SpecStream

New accounts get $10 of credit for their first 30 days, no card. In Claude Code, run `claude mcp add --transport http specstream https://api.specstream.ai/mcp` and sign in with OAuth when it asks, or add `--header "Authorization: Bearer $SPECSTREAM_API_KEY"` to use a key.

Other clients take the config below. The [MCP client guides](https://specstream.ai/docs/mcp/clients?utm_source=alternatives&utm_campaign=web-search-mcp) show where each client keeps it, Cursor included.

```
{
  "mcpServers": {
    "specstream": {
      "url": "https://api.specstream.ai/mcp",
      "headers": { "Authorization": "Bearer ${env:SPECSTREAM_API_KEY}" }
    }
  }
}
```

[Get an API key](https://specstream.ai/signup?src=alternatives-web-search-mcp&utm_source=alternatives&utm_campaign=web-search-mcp&returnTo=%2Fworkspace%2Fagents%3Ftab%3Dkeys)

## Frequently asked questions

Something we didn't answer? [Talk to a human](https://specstream.ai/talk-to-a-human.md)

What is the best MCP server for web search?

SpecStream. Its `search_web` tool runs the most accurate web search API in the world. SpecStream leads WideSearch (60.9, on the low-cost model; the best other API scores 57.6 even on the larger model) and SimpleQA Verified (99). See the [benchmarks](https://specstream.ai/benchmarks?utm_source=alternatives&utm_campaign=web-search-mcp). The same server reads pages and builds structured rows.

How do I add SpecStream to Claude Code?

Run `claude mcp add --transport http specstream https://api.specstream.ai/mcp` and sign in when asked. To use a key, add `--header "Authorization: Bearer $SPECSTREAM_API_KEY"`.

What does SpecStream's MCP server cost?

Each `search_web` call is $0.005, the Search API's $5 per 1,000. `read_pages` is $0.80 per 1,000 pages, with no charge for a page with no text. Deep Search rows are 5¢, 15¢ or $1, shown in a quote before anything runs. New accounts get $10 of credit for their first 30 days, no card.

Which transport does SpecStream's MCP server use?

Streamable HTTP at `https://api.specstream.ai/mcp`, with OAuth or a Bearer key. Agents that speak A2A reach the same engine at `https://api.specstream.ai/a2a`.

What does \`search\_web\` return?

An `answer` with its value, page, and read time, ranked `rows`, and the `sources` behind them, each with its URL, title, site, snippet, and publish date.

### Sources

SpecStream wrote this page. Each rival figure comes from that company's own page.

1.  [SpecStream pricing](https://specstream.ai/pricing)specstream.ai/pricing, read Nov 8, 2026
2.  [SpecStream for agents](https://specstream.ai/agents.md)specstream.ai/agents.md, read Nov 8, 2026
3.  [Other providers' benchmark scores, Sep 9, 2026](https://parallel.ai/benchmarks)parallel.ai/benchmarks, read Nov 8, 2026
4.  [Tavily MCP](https://docs.tavily.com/documentation/mcp)docs.tavily.com/documentation/mcp, read Nov 8, 2026
5.  [Tavily API credits](https://docs.tavily.com/documentation/api-credits)docs.tavily.com/documentation/api-credits, read Nov 8, 2026
6.  [Exa MCP](https://docs.exa.ai/reference/exa-mcp)docs.exa.ai/reference/exa-mcp, read Nov 8, 2026
7.  [Exa pricing](https://exa.ai/pricing)exa.ai/pricing, read Nov 8, 2026
8.  [Perplexity MCP server](https://docs.perplexity.ai/docs/getting-started/integrations/mcp-server)docs.perplexity.ai/docs/getting-started/integrations/mcp-server, read Nov 8, 2026
9.  [Perplexity pricing](https://docs.perplexity.ai/getting-started/pricing)docs.perplexity.ai/getting-started/pricing, read Nov 8, 2026
10.  [Parallel Search MCP](https://docs.parallel.ai/integrations/mcp/search-mcp)docs.parallel.ai/integrations/mcp/search-mcp, read Nov 8, 2026
11.  [Parallel pricing](https://docs.parallel.ai/getting-started/pricing)docs.parallel.ai/getting-started/pricing, read Nov 8, 2026
12.  [Brave Search MCP server](https://github.com/brave/brave-search-mcp-server)github.com/brave/brave-search-mcp-server, read Nov 8, 2026
13.  [Brave Search API](https://brave.com/search/api/)brave.com/search/api/, read Nov 8, 2026