codebase-memory-mcp vs semble
Codebase Memory MCP and Semble both index codebases for agents, but take different shapes: a persistent knowledge graph versus an embedding-based code search.
Side by side
| Dimension | DeusData/codebase-memory-mcp | MinishLab/semble |
|---|---|---|
| Stars | 39,291 | 5,893 |
| Forks | 3,151 | 257 |
| Language | C | Python |
| Last push | 1mo ago | 1mo ago |
| License | MIT | MIT |
| Archived | No | No |
Verdict
Both solve "help my agent understand this codebase" but with different engineering. Codebase Memory MCP builds a persistent knowledge graph with fast, near-zero-latency lookups and broad language support — strong when you need precise symbol-level answers at scale. Semble leans on embeddings and code search, which is lighter to reason about and pairs well with retrieval workflows. If your bottleneck is token efficiency on large repos, Codebase Memory leads. If you want a simpler, embeddable semantic search layer, Semble is easier to reason about. Both are actively maintained, so the deciding factor is graph vs embeddings.
FAQ
Which is faster for large repos?
Codebase Memory MCP is engineered for sub-ms queries against a persistent graph. Semble is embedding-based, which is also fast but indexes differently.
Do they support many languages?
Codebase Memory MCP advertises broad language coverage. Semble focuses on code embeddings that transfer across languages.
Last reviewed 2026-08-06