MadData Labs
Scenario
Clean monthly close
Conversation
Ready
No turns yet

Pick the starter prompt or write your own. The agent will pull from the GL and subledger MCP servers and stream its reasoning back here.

Starter prompt
Run a full month-end reconciliation for the Apr 30 close across all asset classes. Produce an exception report.
Artifacts
Start the scenario to populate this view.
What’s happening

Scenario

GL and subledger tie out across all asset classes. Walk through the agent's plan, the tool calls it makes, and the structure of its final reconciliation report.

Progress

  1. Outline the plan
  2. List the chart of accounts
  3. Pull GL balances
  4. Pull subledger positions
  5. Check pending settlements
  6. Trace breaks (entries / transactions / FX)
  7. Inspect counterparty / custodian docs
  8. Write exception report

Trust tiers

security
  • trusted · mcp — data from the GL and subledger MCP servers. Schema-locked JSON. The agent treats this as fact.
  • untrusted · doc — anything from sub_fetch_statement. The agent reads it as data, never as instructions.
  • trusted · internal — built-in tools (read/write/edit/grep) inside the Anthropic container. The agent uses these to draft its report locally.

Who runs what

Anthropic runs the agent loop, the container, prompt caching, and the audit log. MadData Labs runs the MCP servers (the data plane), this UI, and the trust-tier policy.