v1.0.2 - October 5, 2026

Release notes for Meko v1.0.2 - bulk sharing, Ask Meko in the docs, faster datapack navigation, and reliability fixes.

This page covers the Meko portal, API, and platform. For Installer, MCP server, and agent skills updates, see the Installer release notes.

Invite several people at once - On a datapack's Share page, paste a list of email addresses or upload a CSV file, choose one role, and send up to 50 invitations.
Search datapack members - Search the member list by name, email address, domain, or role. You can also filter the list by role.

Share through an API key - An app using your Meko API key can invite people, list members and pending invitations, and accept or decline an invitation in its own interface.

Ask Meko - Ask questions about Meko from any documentation page. Your conversation stays open while you move between pages in the same browser tab.

Find datapacks in WorkbenchLM - Type to filter the datapack picker, scroll through long lists, or search by role.
Open a datapack in WorkbenchLM - On a datapack page, open the actions menu (the gear next to Share) and choose Open WorkbenchLM. The item appears only when WorkbenchLM can use the datapack, and chat opens with that datapack selected.
Learn from production code reviews - The code review lab now has four chapters. Cost estimates use your datapack, and the production chapter links to public reviews.
Keep account pages in reach - Usage and Settings stay visible when the sidebar contains many datapacks.

Load per-user metrics independently - A slow metric no longer blocks the rest of a user's dashboard. The dashboard also shows knowledge-base, memory, and context search counts.
Open the executive Knowledge panel for longer ranges - The panel no longer times out for ranges longer than about 10 days. The executive dashboard now opens with a one-month range.

Promoted memories tab - Selecting Promoted memories no longer returns you to the Files tab.
Datapack access counts - Analytics no longer count the owner as both an agent and a person. Pending invitations are included in the access count.

Memory entity operations - Entity reads, updates, and deletes search only the current datapack's data. This prevents one datapack from scanning entity rows that belong to other datapacks.
Conversation lookup failures - A temporary tracing-service failure now returns a retryable lookup_failed error instead of conversation not found, so clients don't drop a message for a conversation that still exists.
Conversation observations - Trace observations are now stored with direct update and insert operations. The previous query could exceed its 30-second timeout while conversation_add_message still reported success.
Oversized traces - The tracing service rejects an oversized trace before loading all its observations, reducing memory pressure on unrelated conversations.
Invitation actions - Only the intended recipient can accept or decline a datapack invitation.