> For the complete documentation index, see [llms.txt](https://help.tahua.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://help.tahua.io/product-news/2026/september/admin-read-mcp-assessments.md).

# Assessment read access in tenant admin MCP

## What it does

The hosted, read-only admin MCP (`tahua-admin-read`) lets a connected AI assistant answer questions about an organisation's funding data using tenant-safe, permission-scoped tools. It previously covered funding rounds, funding applications, applicants, and application context — but had no visibility of **assessments** (the reviewer scoring and decision stage). As a result, questions like "summarise the assessments for this round" or "which applications have been recommended for funding" could not be answered.

This feature adds assessment read access to the MCP.

## What's new

Four new read-only tools:

* **`list_assessments`** — lists assessment summaries, filterable by funding round, assessment round, funding application, assessor, or state.
* **`get_assessment`** — one assessment summary by ID.
* **`list_assessment_rounds`** — lists assessment rounds, filterable by funding round.
* **`get_assessment_round`** — one assessment round by ID.

`get_application_context` is also enriched: it now returns the application's **assessments** and its **final assessment score** alongside the existing application, funding round, and applicant context.

### What an assessment summary contains

* Assessment **state** (not started / in progress / complete / declined)
* Aggregate **score** (the cached assessment score), plus **valid-for-scoring** and **conflict-of-interest** flags
* **Decision** (where the round uses decisions) and when it was confirmed
* **Assessor** identity — id, name, and email
* **Assessor type**, whether the assignment was **automatic**, and the due / submitted / published dates
* References to the funding application, funding round, and assessment round

### What it deliberately does not expose

Consistent with the rest of this MCP, assessment tools return **summary data only**. They do **not** expose the raw assessment form answers or per-criterion scores/comments — only the aggregate score and decision. (Per-criterion detail can be added later as a separate, deliberate scope extension.)

## Security and scoping

Assessment access uses exactly the same tenant and data-policy scoping as the existing tools: every query runs through the organisation boundary and the administrator's CanCan data policies (`accessible_by`). An administrator only sees assessments for funding rounds their data policy permits — an admin restricted to some rounds will not see assessments belonging to rounds outside that policy. No new permissions are introduced; the tools inherit the organisation's existing assessment read rules.

## UI changes

None. This is an API/MCP capability only; there are no changes to the admin web interface.
