> SpecStream is the live web data engine for teams and agents. This is the Markdown copy of https://specstream.ai/docs/integrations/atomic-chat. Every page, with what SpecStream does and costs: https://specstream.ai/llms.txt # Atomic Chat Add SpecStream to Atomic Chat as a remote MCP server, and the models in your chats can search the live web, read whole pages and build lists of records. It takes one form on the Connectors page and an API key. [SpecStream for Atomic ChatAI assistants](https://specstream.ai/integrations/atomic-chat.md) ## What you need - Atomic Chat for macOS, Windows or Linux, and a model that can call tools. - A SpecStream account. New accounts get $10 of credit, no card. [Create one](https://specstream.ai/signup?utm_source=atomic-chat). 1. 1 ## 1. Create an API key Atomic Chat's browser sign in covers the connectors in its own list. A server you add yourself sends a header, so SpecStream needs an API key here. In your SpecStream workspace, open [Agents & API](https://specstream.ai/workspace/agents?tab=keys&utm_source=atomic-chat), choose **API keys**, then **Create API key**, and set a total spending limit. Copy the key from the next screen. SpecStream shows it once. 2. 2 ## 2. Open Connectors and add a server In the sidebar, open **Connectors** and select **Add Server**. The **Add MCP Server** form opens. The older **Settings**, **MCP Servers** page now leads to the same Connectors page. 3. 3 ## 3. Fill in the form Enter `specstream` in **Server Name**. Under **Connection**, choose **Web address** and enter `https://api.specstream.ai/mcp` in **URL**. Select **Show advanced options**. Under **Headers**, enter `Authorization` as the **Header name** and `Bearer YOUR_SPECSTREAM_API_KEY` as the **Header value**, with your key in place of `YOUR_SPECSTREAM_API_KEY`. Leave **Transport** on **HTTP (falls back to SSE automatically)** and select **Save**. 4. 4 ## 4. Check that it is connected Saving adds the server and turns it on. The SpecStream card on the Connectors page reads **Connected** when its tools are loaded. If it reads **Couldn't connect**, choose **Edit** and check the key in the header. 5. 5 ## 5. Ask in a chat Start a chat with a model that can call tools, ask for something current and name SpecStream. Unless you allowed all tool calls, Atomic Chat asks you to approve each one: In an Atomic Chat chat ```text Use SpecStream to find the current stable release of Node.js and read its release notes. ``` ## Or add it as JSON In the **Add MCP Server** form, **Add server by JSON** takes the same server as one object, with the server name as its key: ## What Atomic Chat gets SpecStream's server adds these tools to your chats. - `search_web` answers 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_pages` returns 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. `compile` reads the request, `quote_capacity_work` prices it for free, and `authorize_capacity_work` starts it only after you approve the maximum. `get_find` and `list_find_rows` read the rows, each with its source URL and read time. ## Exa and SpecStream side by side Atomic Chat lists Exa as a featured web search connector, with the tools `web_search_exa` and `web_fetch_exa`. With both connectors on, the model chooses between their tools in each turn. To make SpecStream the one web search, switch Exa off on its card, or name SpecStream in your request. ## Approvals and tool switches Under **Advanced** on the Connectors page, **Allow All MCP Tool Permissions** approves every tool call without asking. Leave it off, and Atomic Chat asks before each SpecStream call. To choose single tools, open the menu on the SpecStream card and choose **Tools**. Switching off `authorize_capacity_work` keeps paid Deep Search work from starting in Atomic Chat, while search and page reading keep working. SpecStream keeps its own gate as well: nothing paid starts until you approve the price it quotes. ## Tool call timeout Atomic Chat waits 30 seconds for a tool's answer by default. `read_pages` gives each page up to 30 seconds, so a slow page can reach that limit. Under **Advanced**, set **Tool call timeout (seconds)** to 60. ## Search quality 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. Scores and method are on the [benchmarks page](https://specstream.ai/benchmarks.md). ## Cost Connecting SpecStream to Atomic Chat is free, and new accounts start with $10 of credit, no card. Searches and page reads come out of your balance per call, with no quote. A Deep Search row costs 5¢, 15¢ or $1, depending on what the row needs, never more than the maximum you approved, and a run that finds nothing costs nothing. [Balance and billing](https://specstream.ai/docs/billing.md) has the details. ## Remove Open the menu on the SpecStream card and choose **Delete MCP Server**. To stop the key working everywhere, revoke it in **Agents & API**. More assistants are in [Connect your tools](https://specstream.ai/docs/integrations.md). Questions go to [support@specstream.ai](mailto:support@specstream.ai). ## More [SpecStream for Atomic Chat](https://specstream.ai/integrations/atomic-chat.md) covers what it does in Atomic Chat on one page. Every other tool has its guide in [Connect your tools](https://specstream.ai/docs/integrations.md). You see the most a request can cost before it starts, and the matches are paid from your balance. [Billing](https://specstream.ai/docs/billing.md) has the details. For help, email [support@specstream.ai](mailto:support@specstream.ai).