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

# uni-api

Add SpecStream to uni-api's api.yaml as a search provider, beside or in place of Jina and Tavily, and uni-api's /v1/search endpoint answers from the live web through the SpecStream Search API.

[SpecStream for uni-apiAgent frameworks & builders](https://specstream.ai/integrations/uni-api.md)

## What you need

-   uni-api with an `api.yaml` you can edit, mounted in the container or loaded from `CONFIG_URL`.
-   A SpecStream API key. New accounts start with $10 of credit, no card. [Create an account](https://specstream.ai/signup?utm_source=uni-api).

1.  1

    ## 1. Get a SpecStream key

    [Get your key](https://specstream.ai/workspace/search?route=tavily&utm_source=uni-api&utm_medium=howto) on the Search API screen of your SpecStream workspace.

2.  2

    ## 2. Add the provider to api.yaml

    uni-api sends Tavily's search request to every search provider that is not Jina. SpecStream answers Tavily's request at its own address, so the provider needs only that address and your key. Add it with the `search` model, and allow `specstream/search` on each uni-api key that may search:

    api.yaml

    ```yaml
    providers:
      - provider: specstream
        base_url: https://api.specstream.ai/tavily/search
        api:
          - YOUR_SPECSTREAM_KEY
        model:
          - search
        preferences:
          search_defaults:
            max_results: 10
    
    api_keys:
      - api: sk-xxx
        model:
          - specstream/search
    ```

3.  3

    ## 3. Give it its own name

    Keep the provider name `specstream`, or any name other than `jina`. A provider named `jina`, or one whose `base_url` contains `api.jina.ai`, always sends its searches to Jina's own address. Your other providers and keys stay as they are.

4.  4

    ## 4. Restart and search

    Restart the container so uni-api reads the new file, then search with `GET /v1/search` and the question in `q`. uni-api answers search only by GET: a POST gets `Method not allowed`.

    Terminal

    ```bash
    docker restart uni-api
    curl 'http://localhost:8001/v1/search?q=uni-api%20release%20notes' \
      --header 'Authorization: Bearer sk-xxx'
    ```

## What uni-api sends

`POST https://api.specstream.ai/tavily/search` with your SpecStream key as a Bearer token and Tavily's body: `query`, `topic` `general`, `search_depth` `basic`, `chunks_per_source` 3 and `max_results` 7, or the values in `search_defaults`. The Search API answers in Tavily's shape, `results[].title`, `url` and `content`, and uni-api returns them as `data[]` with `title`, `url`, `description` and `content`.

We ran this file in the `yym68686/uni-api:latest` image on November 12, 2026 and read the request it sent: the address, the Bearer key and the body above, with `search_defaults` applied.

## Settings in search\_defaults

Set these under `preferences`, `search_defaults` on the provider. They apply to every search the provider takes.

| Setting | What SpecStream does |
| --- | --- |
| `max_results` | Returns up to that many results. Up to 10 is one search at the base price. |
| `topic: news` | Runs a news search: dated items, newest first, at the web search price. |
| `topic: general` | Runs a web search. This is the default. |

## Next to Jina and Tavily

uni-api picks a search provider from the ones the key allows. To send every search to SpecStream, list only `specstream/search` under that key's `model`. To keep Jina or Tavily for some keys, give those keys `jina/search` or `tavily/search` instead.

The SpecStream Search API is the most accurate web search API on SimpleQA Verified and WideSearch. The scores and the method are on [the benchmarks page](https://specstream.ai/benchmarks.md). Moving a whole product off Tavily is on [the Tavily migration page](https://specstream.ai/migrate/tavily.md).

## Gemini and Claude search

Searches that a model makes for itself, through the `-search` model suffix with `google_search`, or Anthropic's `web_search` tool, run inside the model provider and do not go through `/v1/search`. This setting changes only the `/v1/search` endpoint.

## Cost

Each search is one Search API call at $5 per 1,000 searches of up to 10 results. It comes out of your balance, with no quote and no approval. With `max_results` at 10 or fewer, every search is billed at the base price, and a search that returns no results costs nothing. New accounts start with $10 of credit, no card.

## Remove it

Delete the `specstream` provider and `specstream/search` from `api.yaml`, then restart uni-api. To end the key's access, revoke it under **Agents & API** in your SpecStream workspace. Questions go to [support@specstream.ai](mailto:support@specstream.ai).

## More

[SpecStream for uni-api](https://specstream.ai/integrations/uni-api.md) covers what it does in uni-api 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).