Track cumulative MCP attribution across requests and thread history (#47081)

## What changed

- Record each MCP source and its first turn, including connector and plugin identifiers when available. Cover direct and Code Mode calls, including returned errors and results that are not printed.
- Persist cumulative attribution checkpoints and restore them across compaction, resume, and forks, including forks limited to recent turns.
- Attach fresh attribution to Responses API `client_metadata` under `mcp_attribution`, outside model-visible content. Restrict transmission to allowed OpenAI HTTPS destinations for both HTTP and WebSocket requests.
- Report `attribution_error` for legacy history without checkpoints, malformed or conflicting attribution, and serialized attribution exceeding 16 KiB.

## Testing

Add coverage for source deduplication, checkpoint persistence and restoration, recent-turn forks, direct and Code Mode calls, destination filtering, and the serialized size limit.

GitOrigin-RevId: fd285e152ca9e24bfa000debb20013709d67a86e
This commit is contained in:
peilin-openai
2026-09-21 18:19:07 +00:00
committed by copyberry
parent 7cec3e110c
commit ac98537678
28 changed files with 1119 additions and 73 deletions
+12
View File
@@ -64,6 +64,18 @@ pub async fn history_with_tool_call_metadata(
items
}
/// Returns the recorder state used by the next session metadata snapshot.
/// This does not change destination filtering or issue a request.
pub fn mcp_attribution_snapshot(
thread: &crate::CodexThread,
) -> codex_protocol::mcp::McpAttribution {
thread
.session
.services
.executed_tool_calls
.mcp_attribution_snapshot()
}
/// Test-only provider that supplies no user instructions.
#[derive(Debug, Default)]
pub struct EmptyUserInstructionsProvider;