mirror of
https://github.com/BerriAI/litellm.git
synced 2026-09-28 01:32:17 +00:00
* fix(mcp): cap an agent key's tools at what the invoking user and team may call The invoking user's and team's x-litellm-user-id / x-litellm-team-id, echoed back by the agent, already narrowed which MCP servers the agent key could reach, but not which tools on those servers. An agent granted every tool on a server kept them all when acting for a user who may only call a subset. The caller's team and user tool grants now intersect the agent's tool list on each server, mirroring the servers axis, so the headers only ever narrow. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com> * test(mcp): pick the caller principal explicitly instead of getattr in the tool grant stub Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> * fix(mcp): return immutable tool sequences from the agent caller tool ceiling Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --------- Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com> Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> |
||
|---|---|---|
| .. | ||
| mcp_server | ||