> ## Documentation Index
> Fetch the complete documentation index at: https://docs.limitguard.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# September 7, 2026

> MCP tool names corrected, and these docs start generating from the API source

## Kept Current Automatically

The price and tool-name table cells on this site are now produced by a generator that
reads the LimitGuard API's own source instead of transcribing it by hand. This entry is
the first such render, run by hand. Once the sync workflow lands, a pull request will
open against this repository whenever a price or an MCP tool name changes there, with the
affected cells already updated. Until then these pages stay hand-maintained.

<Note>
  The [September 5](/changelog/2026-09-05) corrections were found by reading the docs
  against the live API by hand. The generator is meant to make that sweep unnecessary for
  the values it covers — the prices and the MCP tool names in the tables listed below.
</Note>

### MCP tool names corrected

[Pricing](/pricing#mcp-tool-endpoints) paired two of the five REST mirror paths with
tool names the MCP server does not answer to:

| Endpoint                   | Was           | Is                |
| -------------------------- | ------------- | ----------------- |
| `POST /v1/mcp/trust-score` | `trust_score` | `get_trust_score` |
| `POST /v1/mcp/risk-score`  | `risk_score`  | `get_risk_score`  |

The sample `/.well-known/mcp.json` response in
[AI agent integration](/guides/ai-agent-integration#mcp-model-context-protocol) named the
same two tools the same wrong way. It now matches what that endpoint returns.

The names only matter over the MCP transport — `tools/call` with the old name returns
`Unknown tool`. The REST paths themselves are unchanged, so nothing calling
`/v1/mcp/trust-score` over plain HTTP was affected.

The [September 5](/changelog/2026-09-05) entry lists the five tools with the same two
mistakes in it. It is left as written: a changelog records what was said on a date, and
these pages are now generated with that directory excluded for exactly that reason.

### What is generated

| Page                                                 | Generated from                                                                         |
| ---------------------------------------------------- | -------------------------------------------------------------------------------------- |
| [Pricing](/pricing)                                  | every per-call price, tier price and USDC amount, and the tool behind each REST mirror |
| [Quickstart](/quickstart)                            | the quality-tier prices for `/v1/entity/check`                                         |
| [Authentication](/authentication)                    | the four API key tier upgrade prices                                                   |
| [AI agent integration](/guides/ai-agent-integration) | the quality-tier price matrix                                                          |

Prose is not generated, and neither are the JSON samples — the `/.well-known/mcp.json`
sample corrected above is a hand fix, and tool names quoted outside a table stay
hand-maintained. Only the table cells that quote a number or a tool name are generated,
so an endpoint that gains a price but no row here is reported rather than invented — the
sync says so instead of writing a description nobody wrote.