Deterministic verification for AI-generated analysis: run a spec, inspect or diff a workbook
Deterministic verification for AI-generated analysis: run a spec, inspect or diff a workbook
andon re-checks the numbers in finished analysis — the report your AI agent just drafted, the workbook a colleague "quickly updated" — against the data they came from and against themselves. It does this with arithmetic, not with another LLM: reconciliation, internal consistency, schema contracts and Excel workbook integrity, written down as a small YAML spec and enforced with exit codes.
The toolset is worth reading before you wire it up, because it tells you what the integration is really for:
name — no dropped ordersuses — gulmezeren2-byte/andon@v1run — pip install andon-verifyrepo — https://github.com/gulmezeren2-byte/andonRow — level diff for data sheets, not just cell-by-cellFamily — Checksreconcile — row_count, sum, aggregate, group_sum, keysinternal — total_row, percent_sum, recomputeschema — columns, unique, not_null, allowed_values, date_continuityexcel — integrityplausibility — bounds, new_categories, mean_shiftcode — meaningname — no dropped ordersuses — gulmezeren2-byte/andon@v1run — pip install andon-verifyrepo — https://github.com/gulmezeren2-byte/andonreconcile — row_count, sum, aggregate, group_sum, keysinternal — total_row, percent_sum, recomputeschema — columns, unique, not_null, allowed_values, date_continuityexcel — integrityplausibility — bounds, new_categories, mean_shiftcode — meaningCompare MCP servers by category, or browse hosted MCP servers if you prefer an endpoint over a local process.
Check the linked publisher documentation and your connected client's tool list. Names and availability can vary with the server version and account permissions.
| Tool | What it does |
|---|---|
| name | no dropped orders |
| uses | gulmezeren2-byte/andon@v1 |
| run | pip install andon-verify |
| repo | https://github.com/gulmezeren2-byte/andon |
| reconcile | row_count, sum, aggregate, group_sum, keys |
| internal | total_row, percent_sum, recompute |
| schema | columns, unique, not_null, allowed_values, date_continuity |
| excel | integrity |
| plausibility | bounds, new_categories, mean_shift |
| code | meaning |
{
"mcpServers": {
"andon": {
"command": "uvx",
"args": ["andon-verify"]
}
}
}Add to claude_desktop_config.json, then restart Claude Desktop.
Read, search and edit files inside allowed directories with the Model Context Protocol reference Filesystem server.
Retrieve version-aware library documentation and code examples with Context7 MCP, using resolve-library-id and query-docs.
Use Microsoft’s Playwright MCP server to navigate pages, inspect accessibility snapshots, fill forms and capture browser screenshots.
Connect GitHub MCP to repositories, issues and pull requests with GitHub’s hosted service or a locally run server.
Use the Supabase MCP server to inspect database schemas, query project data and retrieve docs through a project-scoped connection.
Search Linear issues and projects, inspect work context and manage updates through Linear’s hosted MCP server with OAuth.