> SpecStream is the live web data engine for teams and agents. This is the Markdown copy of https://specstream.ai/docs/integrations/grinta. Every page, with what SpecStream does and costs: https://specstream.ai/llms.txt # Grinta Add SpecStream to Grinta as a Streamable HTTP server in settings.json, and the coding agent can search the live web, read pages and bring back structured records mid-task. Paid work waits for your yes on a quoted price. [SpecStream for GrintaCoding agents](https://specstream.ai/integrations/grinta.md) ## What you need - Grinta 1.0.0 or newer, installed with `pipx install grinta`. - A SpecStream account and an API key. New accounts start with $10 of credit, no card. [Create one](https://specstream.ai/signup?utm_source=grinta). 1. 1 ## 1. Create a SpecStream API key Open [Agents & API](https://specstream.ai/workspace/agents?tab=keys&utm_source=grinta) in your SpecStream workspace, choose **API keys**, then **Create API key**. Give the key a total spending limit. Grinta sends it as a Bearer token, so there is no browser sign-in. 2. 2 ## 2. Add SpecStream to settings.json An installed Grinta reads `~/.grinta/settings.json`. A source checkout reads the repository's `settings.json`. Add the `specstream` entry to the `servers` list under `mcp_config`, keep your other keys, and put your key in place of `YOUR_SPECSTREAM_API_KEY`. Set `enabled` under `mcp_config` to `true`. It is the master switch for MCP in Grinta, and `grinta init` writes it as `false`. Both `type` and `transport` must be `shttp`. Grinta's `transport` field defaults to `sse`, and SpecStream answers over Streamable HTTP only, so an entry without `"transport": "shttp"` fails to connect. ~/.grinta/settings.json (the mcp\_config part) ```json { "mcp_config": { "enabled": true, "servers": [ { "name": "specstream", "type": "shttp", "transport": "shttp", "url": "https://api.specstream.ai/mcp", "api_key": "YOUR_SPECSTREAM_API_KEY", "usage_hint": "Live web search, page reading, and lists of records with a source on every value. Paid work starts only after the user approves the quoted price.", "enabled": true } ] } } ``` 3. 3 ## 3. Reload MCP in Grinta Restart Grinta, or press **Ctrl+B** to open **Environment**, open **MCP Servers**, and select **Reload**. SpecStream then shows in the **MCP Servers** panel of the sidebar, and you can switch it on or off from the same list. 4. 4 ## 4. Ask in Agent mode Switch to Agent mode with `/mode` and describe the data and the details to bring back. Grinta calls SpecStream's tools through MCP. In a Grinta session ```text Find the open-source Python libraries for reading PDF tables, with each one's repository, license and latest release date. Use SpecStream. ``` ## Why not the Add MCP Server dialog Grinta's **Add MCP Server** dialog, and the **MCP Servers** tab of its settings menu, take a server name and a command or an HTTPS URL. A URL added there is saved with `"type": "sse"` and no key. SpecStream speaks Streamable HTTP and needs the key, so write the entry in `settings.json` as above. After that, the **MCP Servers** list shows SpecStream, and its **Edit**, **Delete** and **Reload** buttons work on it. On November 12, 2026, a FastMCP client, the library Grinta connects with, reached `https://api.specstream.ai/mcp` over Streamable HTTP with Grinta's header and listed SpecStream's tools. The same client over SSE failed to connect. ## Approvals in Grinta Grinta's `/autonomy` sets what it confirms. At `conservative` it confirms every MCP call. At `balanced`, the default, it confirms high risk actions. SpecStream adds its own gate at every level: nothing paid starts until you approve the price it quotes. ## What Grinta calls A request for records moves through these tools in order. - `compile` reads the request and says what SpecStream will look for. - `quote_capacity_work` prices it. It spends nothing. - `authorize_capacity_work` starts the work after you approve the maximum. - `get_find` follows progress, and `list_find_rows` reads the rows, each value with its source URL and read time. ## One fact or one page For a single answer, Grinta can call `search_web` with the question. Each search is $5 per 1,000 searches of up to 10 results, from your balance, with no quote. To read known pages, `read_pages` uses the Extract API at $0.80 per 1,000 pages, and a page with no text is not charged. ## Keep the key out of git Grinta reads `api_key` as written and does not fill in environment variables in MCP entries. In an installed build the file sits in your home folder. In a source checkout, keep `settings.json` out of your commits, or use a key with a small spending limit. ## Cost Connecting SpecStream to Grinta is free. Reading a request and pricing it cost nothing. Each matching row costs 5¢, 15¢ or $1, depending on how much the row needs, plus 5¢ for a person's email and 15¢ for a phone. A quote lasts 30 minutes, a run can't cost more than the maximum you approved, and a run that finds nothing costs nothing. [Balance, quotes, and billing](https://specstream.ai/docs/billing.md) has the details. ## Disconnect Set the entry's `enabled` to `false`, or remove it, and reload MCP. To end the key itself, revoke it under **API keys** in **Agents & API**. Your saved results stay in your workspace. Questions go to [support@specstream.ai](mailto:support@specstream.ai). ## More [SpecStream for Grinta](https://specstream.ai/integrations/grinta.md) covers what it does in Grinta 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).