Frequently asked questions about the Chemia materials MCP connector
These are short answers about the Chemia materials MCP connector: what it is, who can use it, what uses the daily allowance, where its values come from and where to get help. Each answer links to the page that holds the evidence.
| Questions answered | 11 |
|---|---|
| Topics | What it is, access, the allowance, limits, evidence, help |
| Data as of | 2026-10-04 |
| Server release | 1.5.0, read from the live server on 2026-10-04 |
What is the Chemia materials MCP connector?
It is a server that lets an AI assistant call 28 read-only tools that search and read Chemia's materials corpus, over the Model Context Protocol, after an Elixir sign-in (as of 2026-10-04). The overview explains what it is and what it is not.
Which AI clients work with it?
The clients we have tested are listed, with the date of each test, in the Connection guide. As of 2026-10-04 they are: Claude Code (tested on 2026-10-04). We make no claim about a client that is not listed there.
Do I need an account?
Yes. Signing in with an Elixir account is required, an administrator approves each account, and there is no anonymous access. Access, grants and limits explains how to ask for one.
Can the connector change data?
Every tool is annotated read-only: 28 of 28 (as of 2026-10-04).
Which tools use the daily allowance?
3 tools do: compute_pourbaix_stability, count_materials and search_materials (as of 2026-10-04). A charged call uses one unit of the allowance, which is 200 charged calls per day per caller by default. A call whose reply is a clarification question is refunded. Other tools never count against it.
How much can one response contain?
No response carries more than 50 rows (as of 2026-10-04). get_result_set pages through the results of a search.
Where do the property values come from?
Each stored value says what kind of evidence it rests on, in the evidence_source field: experimental, dft (computed, not measured), ml_predicted or literature. When a property has several stored values, one is returned, chosen by the kind of evidence: experimental first, then DFT, then ML. Any other source type, such as literature, is used only when nothing ranked exists, and a tie goes to the oldest stored row. evidence_source says which kind you got; you do not get the range of values. See get_material_property.
How do I read a result correctly?
Read the notes for the tool you used. We sort them into Convention, Trap and Gap: 64 notes as of 2026-10-04. Reading results lists them all.
Is there a machine-readable version of this documentation?
Yes. Every page has a Markdown copy at the same address with index.md added, and the tool reference is published as JSON.
Where are the privacy policy and the terms of use?
The Website privacy policy and the Website terms of use are linked at the foot of every documentation page.
How do I get help?
Use Contact support. Please include the tool name, the arguments you used, the error text and roughly when it happened. Troubleshooting covers the common problems first.