What's new

Highlights from recent Meko releases. For the full history, see the Changelog.

Meko portal and platform — v1.0.1

• Try Meko in Labs — open Labs and launch Review a PR with AgentK. You build datapack context, watch the review, and choose which memories to keep.
• See more on a datapack's Analytics tab — Analytics now show context size, a memory breakdown, and recent search activity. Your datapacks no longer shows an account-wide efficiency tile, or a context badge of 0 on every card.
• Let WorkbenchLM call your datapack — Reads run immediately, as does saving a memory. Deleting a memory, promoting a memory, or updating the datapack waits until you choose Allow.
• Upgrade to Standard on AWS Marketplace — Choose Settings > Pricing to upgrade a Free account to Standard using AWS Marketplace ($25/month). Finish account setup in Meko after you subscribe, or the plan stays on Free.
• Long conversations accept new messages — Add or delete messages after a conversation's stored history exceeds its size limit. Reading or updating that conversation can still fail while the history is over the limit.

Installer and public skills — v2.2.0

• Kiro IDE and CLI support — one install configures Kiro's shared MCP and skills locations and installs durable capture hooks.
• Capture survives delivery failures — Claude Code, Cursor, Codex, and Kiro keep failed exchanges in a local outbox. npx @yugabytedb/meko-mcp@latest --capture-status shows whether capture is healthy, catching up, retrying, or needs action.
• Codex transcripts now capture correctly — Codex uses codex:<repo> ownership and holds incomplete turns for retry instead of advancing past them.
• Safer memory guidance — agents distinguish failed searches from empty results, keep retrieval evidence bounded, verify writes, and treat memory_get_all as a recent window rather than an export.
• Independent releases — the public skills plugin and MCP server are each now released independently from the installer and Desktop extension.

Keep up