run_export
One run’s content-free export (contract 1.32): the run map as a shareable record, with steps in order, agents, models and providers, tools by name, timings, tokens, cost, outcome, approvals, and guardrail and screening events by type. No prompts, outputs, passages or tool arguments. Each export is audited.
Role: viewer and above (any person, AAL staff, machine key); every call is audited as run_exported in the client’s audit log, with via: mcp.
Acts: no. It only reads; nothing an AI agent calls through MCP approves, pays, or changes a limit.
Inputs
Section titled “Inputs”| Input | Type | Default |
|---|---|---|
run_id_hash |
string | required |
The arguments of a tools/call request:
{ "name": "run_export", "arguments": { "run_id_hash": "<run_id_hash>" }}