agentmemory vs MegaMemory
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 — MCP server that lets a coding agent build and query a typed concept graph of the project — SQLite store, in-process embeddings, merge-conflict resolution and a local web graph explorer.
Both are MCP memory servers for coding agents combining a graph with semantic search; agentmemory auto-captures through hooks, MegaMemory expects the agent to curate concepts deliberately — less noise, more discipline required.
| agentmemory | MegaMemory | |
|---|---|---|
| Stars | 27k | 459 |
| Forks | 2.3k | 41 |
| Language | TypeScript | TypeScript |
| License | Apache-2.0 | MIT |
| Last activity | today | 3 months ago |
| Topics | memory, coding | memory, knowledge-graphs, coding |
| Curated connections | 6 | 5 |
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.
MegaMemory — the curator's take
The memory that stores design intent rather than code structure: the agent writes `feature`/`module`/`pattern`/`decision` concepts with typed links, and `understand` does semantic search over them. One npm install, a ~23MB embedding model, `.megamemory/knowledge.db` in the repo, nothing leaves the machine, installers for Claude Code, Codex, OpenCode and Antigravity. Its weakness is inherent: the graph is only as true as the last agent that wrote it — hence the merge/conflict tooling — so treat the DB as regenerable and pair it with an AST-derived index rather than replacing one. Small, single-maintainer project; fine for a repo you own, thin for a fleet.