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

# WalletWall MCP

> A local, read-only MCP server that evaluates stablecoin evidence you supply under published WalletWall policies and reports how well that evidence supports each claim.

WalletWall MCP is a [Model Context Protocol](https://modelcontextprotocol.io) server for one narrow job:
**evaluate structured stablecoin evidence without pretending that evidence is truth**.

Put another way: **WalletWall evaluates evidence; it does not determine truth.**

It runs on your machine. Your MCP client starts it over **stdio**; it needs no WalletWall account, API
key or WalletWall-hosted service. It is **installable today as pre-release software**.

> [!WARNING]
>
> **INSTALLABLE · LOCAL · PRE-RELEASE**
>
> **Product boundary.** WalletWall MCP is local, read-only evidence infrastructure. It does not sign,
> custody, trade, broadcast transactions, or certify truth. It is pre-release and has not had an
> independent security review or audit. See [what you can use today](https://docs.walletwall.org/concepts/product-status.md) for
> every surface side by side.

> [card] [Connect WalletWall](https://docs.walletwall.org/mcp/install.md)
>
> Install the local server in GitHub Copilot, Claude Code or another MCP client.

> [card] [Get a result](https://docs.walletwall.org/mcp/quickstart.md)
>
> Run synthetic evidence end to end and see WalletWall refuse to guess when evidence is missing.

## What problem it solves

Answers about stablecoins often blur what was **observed** with what was **assumed**: a ticker symbol
treated as identity, a provider's number treated as canonical chain state, a failed lookup turned into
zero, or two disagreeing readings quietly collapsed into one.

WalletWall keeps those distinctions explicit. It reports which evidence was presented, where it came
from, which policy applied, whether it was fresh and final, and whether other evidence conflicts with
it. When required evidence is missing, the result stays weak instead of being filled in.

## The mental model

| Concept | What it means here |
| - | - |
| **Evidence collection** | You obtain readings with tools you choose. WalletWall MCP fetches nothing. |
| **Observation** | One reading from one source. |
| **Assertion** | A claim that cites observations or a WalletWall derivation. |
| **Authority** | Which source kind a published policy accepts as able to establish that claim type. Authority is not correctness. |
| **Freshness** | Whether evidence about the present is recent enough at an evaluation time you supply. |
| **Finality** | Whether a chain-derived observation says it came from a finalized block. |
| **Assessment** | How well the presented evidence satisfies the published policy. |
| **Truth** | Whether the underlying facts are actually so. WalletWall does not determine this. |

> [!NOTE]
>
> `ESTABLISHED` means **the presented evidence satisfies the published policy for this claim type**. It
> does not mean WalletWall proved the underlying fact true.

## How it fits together

```mermaid theme={null}
flowchart LR
    accTitle: WalletWall MCP data flow
    accDescr: Evidence you gathered reaches your MCP client. The client calls WalletWall MCP over stdio on your machine. The server evaluates the evidence against its bundled registry, policies, rules and specification, and returns structured results to the client. The client exchanges prompts and tool data with your model provider; WalletWall MCP itself has no network access.

    evidence["Evidence you gathered<br/>chain reads, issuer pages, provider data"]
    client["Your MCP client<br/>GitHub Copilot, Claude Code, others"]
    model["Your model provider"]
    subgraph local["Your machine"]
        server["WalletWall MCP<br/>stdio server, read-only"]
        bundled[("Bundled registry, policies,<br/>rules and WWEP spec")]
    end
    result["Structured results<br/>states, authority, limitations,<br/>conflicts, provenance, digests"]

    evidence --> client
    client -->|"tool call over stdio"| server
    bundled --> server
    server --> result
    result --> client
    client <-->|"prompts and tool data"| model

    class evidence input;
    class client,model external;
    class server process;
    class bundled datastore;
    class result output;

    classDef input fill:#FAF8F3,stroke:#C9A47A,color:#1E1A14,stroke-width:1.5px;
    classDef process fill:#BF4E32,stroke:#8B3120,color:#FAF8F3,stroke-width:1.5px;
    classDef decision fill:#C9A47A,stroke:#8B6F47,color:#1E1A14,stroke-width:1.5px;
    classDef output fill:#E6DED2,stroke:#9A9186,color:#1E1A14,stroke-width:1.5px;
    classDef risk fill:#8B3120,stroke:#1E1A14,color:#FAF8F3,stroke-width:1.5px;
    classDef datastore fill:#1E1A14,stroke:#9A9186,color:#FAF8F3,stroke-width:1.5px;
    classDef external fill:#FAF8F3,stroke:#9A9186,color:#1E1A14,stroke-width:1.5px;
    classDef security fill:#1E1A14,stroke:#C9A47A,color:#FAF8F3,stroke-width:1.5px;
```

*The server is a pure function of the bundle you send, the evaluation time you send, and the registry,
policies and rules bundled into it. It opens no network connection, writes no files, reads no
environment variables and keeps no query history. Your MCP client and model provider do see every tool
input and output: see [Privacy & Security](https://docs.walletwall.org/mcp/privacy-security.md).*

## What ships

* **Six tools** for registry lookup, normalization, evidence evaluation, invariant checks, provenance and
  vault technical-readiness assessment. See [Tools](https://docs.walletwall.org/mcp/tools.md).
* **Eleven resources** containing the evidence contract, policies, registry, versions and synthetic
  examples. See [Resources](https://docs.walletwall.org/mcp/resources.md).
* **WWEP 0.1 (draft)**, the WalletWall Evidence Profile behind every result. See
  [Evidence Model (WWEP)](https://docs.walletwall.org/mcp/evidence-model.md).
* **GitHub Copilot and Claude Code integrations**, both running the engine locally.
* **WallPrint 0.1**, a separate disclosure contract and public resolver for records WalletWall
  deliberately publishes. Calling MCP never publishes a WallPrint.

## Installable today — pre-release

| Surface | Status |
| - | - |
| WalletWall MCP stdio server | **Installable today — pre-release.** Six tools and eleven resources; source in `walletwall-copilot`. |
| GitHub Copilot plugin | **Installable today — pre-release.** The GitHub repository acts as its plugin marketplace. |
| Claude Code local plugin | **Installable today — pre-release.** Distributed from [`Wallet-Wall/walletwall-claude`](https://github.com/Wallet-Wall/walletwall-claude). |
| Official MCP Registry entry | Listed as `io.github.Wallet-Wall/walletwall`; this is discovery metadata, not endorsement or a maturity signal. |
| Loopback HTTP adapter | Optional pre-release companion. You run it yourself on `127.0.0.1`; WalletWall does not host it. |
| WallPrint public resolver | **Available.** Hosted at `walletwall.org/wallprint`; records appear only after deliberate WalletWall publication. |

## Not part of the product

These capabilities are deliberately absent from WalletWall MCP, and no page in this section describes
them as available.

* **No WalletWall-hosted MCP endpoint, remote transport, managed agent or hosted inference.**
* **No live data retrieval inside the engine.** You supply the evidence; WalletWall evaluates it.
* **No transaction capability of any kind**: no keys, signing, custody, trading, transaction
  construction, broadcasting or vault execution.
* **No signatures or attestations on WallPrint records.** A digest proves bytes, not authorship or
  factual truth.
* **No certification.** Results do not certify regulatory compliance, solvency, safety or audit status.

## Status

WalletWall MCP is **pre-release**. WWEP 0.1 and WallPrint 0.1 are WalletWall-defined drafts, not
industry standards, and neither the specification nor the code has had an independent security review
or audit. Version domains move independently; [Install & Connect](https://docs.walletwall.org/mcp/install.md) shows how to inspect the
server you are actually running.

## How this relates to the WalletWall application

The hosted application at walletwall.org and the MCP server share a discipline, not a runtime. The
application fetches data from providers and scheduled snapshots and labels every response with its
source; the MCP server fetches nothing and evaluates the evidence you present. The application's
[Wallet Evidence Model](https://docs.walletwall.org/architecture/wallet-evidence-model.md) and the server's
[Evidence Profile](https://docs.walletwall.org/mcp/evidence-model.md) are different schemas built on the same rule — unknown is never
zero, and a source's authority is never inferred from how its bytes arrived. The same rule governs how
WalletWall reasons about [independent authority](https://docs.walletwall.org/architecture/independent-authority.md) in its own
assurance work.

> [card] [Data Serving Architecture](https://docs.walletwall.org/architecture/data-serving-architecture.md)
>
> How the hosted application sources and labels its own data, for contrast with the bring-your-own-evidence model here.

> [card] [What you can use today](https://docs.walletwall.org/concepts/product-status.md)
>
> Where the MCP server, the plugins and WallPrint sit among WalletWall's surfaces by maturity.

## Where to go next

> [card] [Install & Connect](https://docs.walletwall.org/mcp/install.md)
>
> Connect the local server to your MCP client.

> [card] [Quickstart](https://docs.walletwall.org/mcp/quickstart.md)
>
> Run a synthetic example and see the fail-closed behavior first.

> [card] [Evidence Model (WWEP)](https://docs.walletwall.org/mcp/evidence-model.md)
>
> Understand sources, observations, assertions and evidence states before diving into the reference.

> [card] [WallPrint](https://docs.walletwall.org/mcp/wallprint.md)
>
> Learn where local evaluation ends and deliberate publication begins.

> [card] [Tools](https://docs.walletwall.org/mcp/tools.md)
>
> Reference for the six read-only tools.

> [card] [Resources](https://docs.walletwall.org/mcp/resources.md)
>
> Reference for the eleven static resources.

> [card] [GitHub Copilot](https://docs.walletwall.org/mcp/github-copilot.md)
>
> Install the plugin, use its four skills and open the WalletWall canvas.

> [card] [Claude / local HTTP adapter](https://docs.walletwall.org/mcp/claude-local-adapter.md)
>
> Use the Claude Code plugin or the optional user-run loopback adapter.

> [card] [Privacy & Security](https://docs.walletwall.org/mcp/privacy-security.md)
>
> Understand exactly what runs where and what your model provider can see.

> [card] [Troubleshooting](https://docs.walletwall.org/mcp/troubleshooting.md)
>
> Diagnose setup, evidence and resolver problems by symptom.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.