Claude-Mem explained: what a coding agent remembers
Deploying Claude-Mem: local worker and server API are different
Plan runtime dependencies, storage, access control and recovery for the path you choose
What you will learn
- Check runtime requirements
- Choose local or remote
- Prepare operations
Before you start
- A synthetic two-session project
- One supported host
- A deliberate memory provider choice
Use two synthetic sessions and an explicit deletion check before team adoption
Key takeaways
- The package engine is stricter than the README summary.
- Remote API controls require configuration.
- Storage needs an explicit recovery plan.
Check runtime requirements
The pinned README lists Node 20 and describes Bun, uv and bundled SQLite. `package.json` is more specific: Node `>=20.12.0` and Bun `>=1.1.31`. Use the stricter package requirement for a reproducible install.
The local architecture document shows a per-user HTTP worker, SQLite and Chroma storage. Check its configured bind address and data directory before exposing any interface beyond loopback.
Choose local or remote
The server API document describes `/v1` endpoints, optional API-key auth and a read-only remote MCP recall endpoint. Its beta server path should not be mistaken for the local worker API.
Rate limits, quotas and metering in that document are opt-in; unset controls are not evidence of enforced production limits. A public deployment needs explicit authentication, network boundary, backup and retention design.
Prepare operations
Record the selected host adapter, provider, package revision, worker status, storage paths and restore procedure. The production guide lists queue depth, active sessions, WAL and Chroma growth as health signals.
We did not start a worker, deploy the server API or test a restore. The guide’s historical throughput and savings numbers are upstream experience, not results reproduced here.
Decision guide
| Criterion | Option A | Option B |
|---|---|---|
| Best when | You need predictable behavior and easy auditing | You need adaptive optimization and have reliable telemetry |
| Main risk | May leave performance on the table | Can become difficult to explain or debug |
Implementation steps
- 1
Meet the package engine requirements.
- 2
Keep local worker and server API boundaries separate.
- 3
Test health, backup and restore before broader access.
Copy-ready example
runtime: node_20_12_plus_and_supported_bun
mode: local_worker_or_server_api
storage: backed_up_and_access_restricted
auth: explicit_for_remote_access
restore: testedFrequently asked questions
Is the worker safe to expose publicly by default?
Do not infer that from a local setup guide; verify bind address and access controls first.
Does the local plugin require the beta server API?
The documents describe them as different runtime paths. Choose the integration actually needed.
Sources
- Claude-Mem / README.mdSource checked 2026-10-08
- Claude-Mem / package.jsonSource checked 2026-10-08
- Claude-Mem / docs/architecture-overview.mdSource checked 2026-10-08
- Claude-Mem / docs/production-guide.mdSource checked 2026-10-08
- Claude-Mem / docs/api.mdSource checked 2026-10-08