# Local AI memory vs hosted APIs: what changes?

> Compare local project storage, retrieved context and optional hosted services before selecting an AI memory workflow for project documents.

- URL: https://neuroaigent.ai/central-brain-answers/local-memory-vs-hosted-apis/
- Updated: 2026-10-18

**A local project-memory workflow keeps its default preparation, index and memory on storage you choose; a hosted service receives the data sent to it.** Central Brain separates its default local workflow from optional cloud connections. A connected cloud AI client still receives the context it retrieves, so local storage and provider processing must be assessed separately.

Choose a workflow by tracing where each part of the project goes: originals, prepared output, vectors, memory and retrieved excerpts. A label such as local-first cannot answer every storage or retention question. Central Brain documents its default behavior and optional connections in its [security guide](https://neuroaigent.ai/central-brain/security/).

A bundled 384-dimension MiniLM model and a local JSONL index; cloud embeddings are optional. Start by identifying what the default local setup supplies before adding an external component. Evaluate any hosted component against an actual project requirement, the material it receives and who administers it.

## Separate storage from processing

A vector index, a note store and an assistant are different parts of the workflow. Review each configured service separately. Do not infer that an external service has the same deletion or backup behavior as your local project folders.

| Layer | Default local workflow | Question for a hosted component |
| --- | --- | --- |
| Project files and prepared output | Stored on storage you choose. | Which files or excerpts will be sent, and for what purpose? |
| Embeddings and retrieval index | Bundled model and local index. | Where are vectors and associated text stored? |
| Project memory | Readable local notes and brief. | Will any copy be stored or processed externally? |
| Connected cloud assistant | Receives the context returned by its tools. | Which provider and workspace settings govern that context? |

## Review these boundaries before connecting project material

1. List the selected folders, prepared-output location, memory store and backups. Identify who can access each.
2. Inspect the enabled connections. Record their purpose and the selected data they receive.
3. Identify the AI client and provider settings that will govern retrieved excerpts. Confirm the project permits that processing.
4. Check a known search result and open its source. Confirm that the selected project excludes unrelated material.
5. Define a retention and recovery process for local folders and each enabled external service.

## Pilot the default workflow first

Use a non-sensitive collection with documents you know. Prepare and search it locally, then connect one supported client. Inspect the returned context and compare it with the original source. A successful connection establishes access; human review still determines whether the returned material is suitable for the task.

Add optional hosted components only after naming the need they solve. Repeat the data review when the connection changes. Keep a short configuration record so another project owner can understand which boundary was approved and reproduce the checks.

## Account for the whole operating environment

In the default workflow, files, prepared output, vectors and memory stay on storage you choose. Keygen receives licensing and device data, not documents. Connected AI tools receive the context they retrieve, and optional cloud connections send selected data only when enabled. Synced folders, device permissions and backups also affect who can access local material. Review them as part of the deployment.

Central Brain runs air-gapped: preparing files, indexing, search and memory need no network connection. The license is re-checked online every 14 days, and Team and Enterprise plans can get a build with no online license check. Optional cloud connections and cloud-hosted AI clients need internet. Distinguish those operating requirements from where documents are stored. Local preparation and a cloud-assisted answer can be parts of the same workflow, with different network and processing boundaries.

## Common questions

### Does local indexing mean no project context reaches a cloud AI provider?

No. A connected cloud assistant receives the context it retrieves. Review its provider and workspace settings as well as the local project configuration.

### Are cloud embeddings required for the default Central Brain workflow?

No. The default workflow uses a bundled embedding model and local index. Optional cloud connections introduce their own selected-data boundaries.

### What should I review when adding a hosted service?

Document the data sent, storage location, access controls, retention, deletion and backup expectations. Review each configured service separately.

## Related answers

- [One memory for ChatGPT, Claude and Copilot](https://neuroaigent.ai/central-brain-answers/one-memory-chatgpt-claude-copilot/): Use Central Brain as one local project memory for supported ChatGPT desktop, Claude and GitHub Copilot clients. Set up and verify cross-client recall.
- [Keep AI project notes in plain files](https://neuroaigent.ai/central-brain-answers/plain-files-ai-project-notes/): Build a readable project brief and decision log, then use Central Brain checkpoints and source-traceable retrieval to reuse approved context.
- [ChatGPT memory vs a local AI memory app](https://neuroaigent.ai/central-brain-answers/chatgpt-memory-vs-local-memory-app/): Compare ChatGPT native memory with Central Brain local project memory: scope, file search, portability, control and what context reaches a cloud assistant.

## Next steps

- [Review Central Brain plans](https://neuroaigent.ai/central-brain/pricing/)
- [Start a local project](https://neuroaigent.ai/central-brain/getting-started/)

## Sources and review notes

- [Central Brain security and data boundaries](https://neuroaigent.ai/central-brain/security/)
- [Central Brain local project workflow](https://neuroaigent.ai/central-brain/)
- [Central Brain MCP memory server](https://neuroaigent.ai/mcp-memory-server/)

## Explore the complete Central Brain workflow

Bring local file preparation, source-traceable search and readable project memory together. Explore [Central Brain features](https://neuroaigent.ai/central-brain/), follow the setup guide and choose a plan for your project.