Is MCP available in Innguma?

What does it mean that it is available?

It means the connection is built and working: it is not a pilot or an announced feature. An MCP server publishes a set of tools that the assistant discovers and can invoke; the specification allows that set to depend on the authorisation with which each client connects^[1]. At Innguma this comes down to a simple principle: Innguma defines which knowledge and which capabilities it makes available to each integration, and your organisation decides which ones it enables.

What can you do from your assistant, stage by stage?

With the capabilities your organisation enables, the connected assistant can take part in the whole watch cycle:

1. Sources. Organise and add information sources, both from Innguma’s source library and new ones, and audit their performance with suggestions for improvement.
2. Filters. Modify the filters that define each watch.
3. Review. Consult the news in the inbox and ask for whatever you need about it: summarise, compare, prioritise or prepare a briefing.
4. Publication. Publish content automatically, with summaries, insights, tags or additional fields, and edit or delete news items already published.
5. Query. Search what has been published to answer a question with the history of your watch.

An illustrative example: ‘Summarise this week’s signals on solid-state batteries, group those that refer to the same announcement, and prepare a draft note for executive management with the links to each source’. Your AI provides analytical capacity. Innguma provides context.

What does your organisation decide before enabling it?

More capabilities are not always better. OWASP places excessive agency among the main risks of applications built on language models, and recommends offering the agent only the functions it needs^[2]. The Spanish Data Protection Agency (AEPD), in its February 2026 guidance on agentic AI, notes that granting agents excessive permissions is a critical risk factor^[3]. Before activating the integration, it is worth deciding:

• Which capabilities are enabled: query only, or also management of sources, filters and publication.
• Which actions may be carried out automatically, and which require a person to review them first.
• Which assistant the organisation will use, and who administers it.
• Who answers for the content that is distributed.

How do you get started?

1. Say which assistant your organisation already uses: ChatGPT, Claude, Microsoft Copilot or Gemini.
2. Choose two or three real tasks to try, for example a weekly summary or a review of sources that are not contributing.
3. Decide which capabilities to enable at the start. Beginning with querying and extending later is a reasonable practice.
4. Define who reviews what the assistant produces before it reaches other people.

The access conditions for the integration are set with each organisation during set-up. On the assistant’s side, each provider determines which plans allow MCP connectors to be added: check its official documentation^[4].