Fetching from the wire…
Public story · 2026-09-18 · high
Mem0 drops whole caller entries when the header stack fills, never slicing names. That stops a cut identifier from being read as a different client.
Why now: mem0 shipped 2.1.0 on September 18.
mem0 released version 2.1.0 with a specific fix to how it identifies callers. mem0's v2.1.0 release notes add three set-once headers that carry surface identity. One of them, X-Mem0-Client, is append-only and lists name/version for every layer in the call chain, outermost first.
That matters if you've built on an SDK that gets wrapped by something else. Before this, a plugin calling into the mem0 SDK reported only itself as the client, not the plugin that invoked it. Now the whole chain shows up, so debugging which layer in a stack made a call gets you the full path rather than the last hop alone.
The header stack has a size limit. When it's full, mem0 drops whole entries instead of slicing characters off the end. Cutting a name mid-string could leave a fragment that looks like a different, real client name. The platform would then read that fragment as a legitimate one.
Each link below shares sources, entities, or timing with this story.
If you wrote an MCP server before July, it's on a protocol shape the maintainers have already removed. Not deprecated-with-a-migration-window. Removed from the spec. MCP lead maintainers David Soria Parra and Den Delimarsky published an updated roadmap on August 22, and the re...
I check Product Hunt maybe once a week and usually regret it. Today's board is worth reading as market structure. The July 30 leaderboard: SKI at 277 upvotes (free voice input for Claude Code and Codex). AI Search Console at 249 (prompt analytics and citation mapping). Memmy A...
Memori turns agent execution traces into structured persistent state, outperforming Zep, LangMem, and Mem0 on the LoCoMo benchmark while reducing prompt size by 67% vs Zep. Python and TypeScript SDKs. If you're building agents that need memory, benchmark this against whatever...
MemPalace (57,821 stars, v3.6.0) reports 96.6% raw recall@5 on LongMemEval with no LLM required, 98.4% with hybrid v4 on a held-out 450 questions, LoCoMo R@10 rising 60.3% → 88.9%, ConvoMem 92.9%, MemBench 80.3%, while explicitly refusing head-to-head comparison against Mem0,...
Three breaking changes: A2A task state tracking and updating, clarified A2A agent run modes, and file_access_read_lines with the line-numbering contract relocated. Dependency work removes the Azure.AI.OpenAI dependency outright and swaps the deprecated AWSSDK.Extensions.Bedroc...
Blaizzy/nativ (1,163 stars, Swift, MIT, macOS 26+) comes from the mlx-vlm author and bundles that server into a SwiftUI app that discovers MLX models already in your HF cache. It exposes OpenAI-compatible chat, Responses, image, audio and model endpoints plus Anthropic Message...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.