> For the complete documentation index, see [llms.txt](https://docs.talordata.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.talordata.com/serp-api/integration/mcp-integration/hermes-agent-integration/hermes-plugin-integration.md).

# Hermes Plugin Integration

Install the TalorData plugin in Hermes Agent to give your agent live Google search backed by the TalorData SERP API. The plugin supports everyday web research through Hermes's built-in `web_search` workflow, as well as advanced Google searches that need specific locations, languages, result types, time ranges, or pagination.

## Prerequisites

Before you begin, make sure that you have:

* A TalorData account with an active API Token for the SERP API.
* A working Hermes Agent installation.
* Access to the Hermes Agent Python environment.
* `uv` available in your terminal.

If you are new to Hermes Agent, complete its installation and quickstart before installing the plugin.

{% stepper %}
{% step %}

## Get your TalorData API Token

Sign in to the [TalorData dashboard](https://dashboard.talordata.com/) and create or copy an API Token.

{% hint style="warning" %}
Keep this token private. Hermes uses it to authenticate every TalorData search request, and those requests consume quota from your TalorData SERP API plan.
{% endhint %}
{% endstep %}

{% step %}

## Install the TalorData plugin

Clone or download the [talor-hermes-plugin repository](https://github.com/TalorData/talor-hermes-plugin). Then install the package into the Python environment used by Hermes Agent, replacing `<path-to-repo>` with the local path to that repository.

{% tabs %}
{% tab title="macOS, Linux, or WSL2" %}

```bash
cd ~/.hermes/hermes-agent
uv pip install --python venv/bin/python <path-to-repo>
```

{% endtab %}

{% tab title="Native Windows (PowerShell)" %}

```powershell
cd $env:USERPROFILE\.hermes\hermes-agent
uv pip install --python .\venv\Scripts\python.exe <path-to-repo>
```

{% endtab %}
{% endtabs %}

If you run Hermes from Git Bash, use forward slashes instead: `cd ~/.hermes/hermes-agent` and `--python venv/Scripts/python.exe`. If you have set `HERMES_HOME`, use the Hermes Agent directory under that location instead.

{% hint style="warning" %}
The plugin must be installed into the same Python environment that launches Hermes. Installing it into another Python environment will prevent Hermes from discovering it.
{% endhint %}
{% endstep %}

{% step %}

## Enable and configure the plugin

Enable the plugin, then inspect the tools Hermes can discover:

```bash
hermes plugins enable talor
hermes tools
```

When Hermes prompts you to configure web-search providers, select **TalorData** under **Web Search & Extract** and enter the API Token from the TalorData dashboard. The plugin uses the `TALOR_API_TOKEN` setting for all TalorData search requests.

If Hermes asks whether the plugin may replace built-in tools, choose **no**. The plugin adds the TalorData web-search provider and `talor_google_search`; it does not replace Hermes's existing tool handlers.

Restart any Hermes Agent session that was already running so it can load the enabled plugin and configuration.
{% endstep %}

{% step %}

## Verify the installation

Start a new Hermes Agent session and send a request such as:

```
Find the five most recent news articles about AI agents.
Return the title, publisher, publication time, source URL, and a one-sentence summary for each result.
```

When the installation is working, Hermes should use its `web_search` workflow with TalorData as the backend and return live results with source URLs.

To check the advanced Google search tool, try a request with explicit search conditions:

```
Search Google News for OpenAI stories from the past week.
Use United States results and English language settings. Group the results by topic and include the original article URLs.
```

Hermes should use `talor_google_search` when the request needs location, language, search-type, time-range, device, or pagination controls. Tool selection is based on the task, so verify the selected tool and response shape rather than expecting identical listings over time.
{% endstep %}
{% endstepper %}

## Example prompts

### Market research

```
Compare the five leading AI meeting assistants in the United States.
Include each company's positioning, pricing signals, and source links.
```

### Location-aware product research

```
Compare the best budget Android phones currently available in Germany.
Search with German location and language settings, check at least three retailers,
and return a shortlist with current prices and key specifications.
```

### Image search

```
Find official renders of the Xiaomi 15 Ultra in Google Images.
Return the image source, page title, and source URL for each result.
```

### Mobile search results

```
Show mobile Google search results for "best dim sum Singapore".
Include the title, URL, and a short summary for each organic result.
```

## What the TalorData plugin can help Hermes Agent do

The plugin gives Hermes two TalorData-backed Google search paths:

| Request type           | Hermes tool           | Use it for                                          |
| ---------------------- | --------------------- | --------------------------------------------------- |
| General web research   | `web_search`          | Everyday research, discovery, and source collection |
| Advanced Google search | `talor_google_search` | Requests that need explicit search controls         |

For advanced searches, `talor_google_search` supports Google parameters including:

* Device, location, Google domain, country (`gl`), and language (`hl`)
* Search types through `tbm`, including images, shopping, news, videos, local results, and patents
* Time ranges through `tbs`, from the past hour to the past year
* Safe-search settings, result pagination (`start` and `num`), JavaScript rendering, AI Overview, and cache controls

The plugin returns focused search fields for Hermes rather than passing a complete raw SERP response into the agent context. Ask Hermes to preserve source URLs and to state the search country, language, time range, and other conditions when they affect the result.

## Troubleshooting

<details>

<summary>The TalorData plugin does not appear</summary>

Confirm that the package was installed into the Python environment that runs Hermes, then run `hermes plugins` to confirm that `talor` is enabled. Restart Hermes after enabling the plugin.

</details>

<details>

<summary>TalorData does not power web search</summary>

Confirm that TalorData is selected under **Web Search & Extract** and that `TALOR_API_TOKEN` is available to the Hermes process. Start a new session after changing either setting.

</details>

<details>

<summary>Authentication or quota error</summary>

Check that the API Token is valid and active in the TalorData dashboard. If necessary, create a new token and update the Hermes configuration. Check your plan and remaining SERP API quota before retrying the request.

</details>

<details>

<summary>The search result is empty or unexpected</summary>

An empty result is not necessarily an installation error. Try a broader query, another country or language, or a different Google search type. Search results are time- and location-sensitive.

</details>

<details>

<summary>Hermes does not use the advanced search tool</summary>

Add the required search conditions to your request, such as a target country, language, time range, search type, device, or pagination. Hermes selects tools from the task rather than from a fixed keyword.

</details>

## Security and usage notes

* Do not paste your TalorData API Token into prompts, shared screenshots, public repositories, or client-side code.
* Use TalorData only for public search result data and comply with applicable laws, platform terms, and your internal data-use policies.
* Preserve source URLs and query conditions in reports that need traceability.

## Next steps

* Use Hermes Agent to build repeatable news briefings, market research, product comparisons, and location-aware search workflows.