agentmemory vs memvid
Persistent memory for coding agents on the iii engine: MCP server with 53 tools, 12 auto-capture hooks, hybrid search + knowledge graph, zero external DBs. Claims 95% R@5 and 92% token cuts. — versus — Single-file memory layer for agents: data, embeddings, index and metadata in one portable .mv2 — append-only Smart Frames, time-travel queries, sub-5ms recall, no server. Rust core, Node/Python SDKs.
Opposite packaging of the same promise: agentmemory is a server with 53 MCP tools, capture hooks and a knowledge graph; memvid is one portable .mv2 file carrying data, embeddings and index. Rich surface vs move-it-anywhere.
| agentmemory | memvid | |
|---|---|---|
| Stars | 27k | 16k |
| Forks | 2.3k | 1.4k |
| Language | TypeScript | Rust |
| License | Apache-2.0 | Apache-2.0 |
| Last activity | today | 1 months ago |
| Topics | memory, coding | memory, local |
| Curated connections | 6 | 3 |
agentmemory — the curator's take
The benchmark-forward entry in a crowded field — '#1 on real-world benchmarks' is self-run, so weigh it accordingly; what's independently real: 1,428 tests, the viral design gist it implements (Karpathy's LLM-wiki pattern plus confidence scoring and lifecycle), and the smoothest onboarding in the category — hand your agent one URL and it installs itself. The iii engine is your infrastructure bet: a server on :3111, not a library. A 53-tool MCP surface is the opposite of the plain-files philosophy — richer, but agents need guidance to use it well (they ship 15 skills for exactly that reason). Same prolific author as pro-workflow and tailclaude — expect fast movement, budget for churn.
memvid — the curator's take
The portability bet is the real differentiator: one .mv2 capsule holds the whole memory — hand it to another agent, ship it offline, branch it, rewind it. The project began life literally encoding text into video frames; the codec DNA survived the Rust rewrite as compression smarts, and the maturity question survives with it. Benchmark claims (+35% SOTA on LoCoMo, '1,372× throughput') are self-run — the eval is open, so run it before repeating it. When NOT: team-shared concurrent memory — single file, append-only, no server means no multi-writer story; that's what the server-based layers are for.