MMESHSYSTEMS

MESH AI PLATFORM / DOCUMENTATION

Agent Chat & Execution Inspector

Use persistent conversations and inspect actual execution evidence.

Agent Chat & Execution Inspector in the actual Mesh AI Portal
Actual authorized Portal capture, 27 October 2026. Retained synthetic support-policy questions and cited answers; no new inference was initiated for photography. Account controls, immutable resource identifiers and private traces are excluded.

Start a conversation

  1. Choose an Agent and an immutable version in Agent Chat.
  2. Start a new conversation and send a bounded request.
  3. Provide missing information when asked.
  4. Expand actual Tool execution records to inspect authorized arguments and sanitized results.
  5. Resume your private conversation from history; refresh restores persisted messages subject to retention and access.

Understand context and evidence

Conversation context is bounded. Separate conversations do not share their history automatically. Repeated wording does not prove memory: the Execution Inspector identifies included history references, immutable versions, model invocation metadata and Tool chronology.

Inspector visibility

Inspection requires the server-authorized permissions and ownership of the conversation. The inspector distinguishes captured metadata, unavailable measurements and redacted resources. It does not expose raw provider prompts, confidential document text or hidden model reasoning.

Current boundary

Agent Chat is separate from Workflow Studio Test Chat. General specialist delegation, embedded delivery and conversational Workflow invocation remain roadmap work.