Editions
The whole engine, on your machine, for free.
This is the same engine we run in production, shipped as one signed binary. No vector database, no search cluster, no API key, and no account needed to run it. Install it into Claude Code, Cursor or Codex with one command and your agent stops forgetting. The free tier limits scale and gates some of the advanced capabilities. Cloud takes over hosting, sync and collaboration, and adds the compliance packs.
The engine is compiled rather than open source. Our SDKs and the benchmark harness are public, so you can reproduce every number we publish against the binary you installed and check the claims without us shipping the source.
Download & run
One command. No Rust toolchain.
The installer grabs a prebuilt binary for your platform and wires MemMesh into every AI tool it detects.
macOS · Linux
curl -fsSL https://memmesh.ai/install.sh | shWindows · PowerShell
irm https://memmesh.ai/install.ps1 | iexPrebuilt and signed for macOS (Apple Silicon and Intel), Linux (x64 and ARM64), and Windows x64. Prefer to grab a binary directly? See the releases →
Then memmesh install adds the MCP server + an auto-observe hook, and restart your AI tool.
Editions
Where the line falls.
You run the same engine either way. What changes is your plan, not the build. Cloud takes over hosting, sync and collaboration, and adds the compliance packs regulated teams need.
| Capability | Local | Cloud |
|---|---|---|
| Durable cross-session memory — typed, bi-temporal | ||
| Runs 100% local — one signed binary, SQLite, no API key, no LLM requiredCloud is managed; self-host/air-gap is an Enterprise option | included* | |
| MCP-native — Claude Code, Cursor, Codex, Windsurf | ||
| Auto-observe — the engine decides what's worth remembering | ||
| Keyword search | ||
| Semantic + hybrid retrieval (vector + keyword + recency) | ||
| Knowledge graph — entities, edges, point-in-time replay | ||
| Basic consolidation / de-duplication | ||
| Cross-encoder reranking (accuracy-tuned retrieval)In the engine; plan limits apply | included* | |
| Multi-hop graph reasoningIn the engine; plan limits apply | included* | |
| Predict any declared target from historyIn the engine; plan limits apply | included* | |
| Calibrated confidence + honest abstentionIn the engine; plan limits apply | included* | |
| Behavior mining — emergent patterns nobody definedIn the engine; plan limits apply | included* | |
| Multimodal ingestion — image, audio, documents | — | |
| Compliance packs — HIPAA / FCRA / EU-AI, GDPR export + erasure | — | |
| Managed cross-device sync + hostingLocal syncs if you run both ends yourself | included* | |
| Support + SLALocal: community support | — |
* Included in the local engine, subject to free-tier limits. Exact limits are still being finalised. See pricing for the current tiers.
Why this split
Free where it earns trust. Cloud where it earns its keep.
A real store, not a teaser
Durable typed memory, bi-temporal replay, a knowledge graph, semantic search, reranking, prediction and MCP wiring. Enough to run a serious agent entirely on your own machine, forever, for free.
Scale is what you pay for
You get the real engine on day one and pay once it becomes load-bearing: more subjects, more forecasts, managed hosting, a team on it, the compliance packs. It isn't a demo that expires.
Yours to check, and to leave with
The engine is compiled, but the SDKs and benchmark harness are public, so you can run our numbers yourself. Your data stays on your machine and exports whenever you want. Cloud is an upgrade you choose because it's better, not one you're stuck with.