Straight answers about Central Brain™.

Central Brain™ is a local-first AI memory layer by NeuroAIgent: a Windows and Apple silicon Mac desktop app that indexes project files locally and gives supported AI tools, such as Claude Desktop and GitHub Copilot, persistent project memory and source-traceable search over MCP (Model Context Protocol). It prepares PDF, DOCX, TXT, Markdown, CSV and XLSX files and searches them with a bundled embedding model. Claude Desktop, Claude Code, GitHub Copilot and the ChatGPT desktop app connect on Windows and Apple silicon Macs. Preparing files, indexing, search and memory run without a cloud AI provider, and licensing uses Keygen. See the product page.

Yes, in the practical sense: it is a store of your files and project notes that you and your AI assistants can both read. The MCP server keeps a per-project memory of five note types (decisions, facts, questions, todos and summaries) plus a project brief, saved as local files. Your assistants can read it at the start of a session and add to it as work happens. Unlike a notes app, the store is searchable by meaning and reachable from Claude Desktop, Claude Code, GitHub Copilot and the ChatGPT desktop app. Unlike built-in chat memory, it lives on storage you choose. Read more at second brain for AI.

Central Brain™ is for people who work with an AI assistant across many sessions on the same projects and are tired of re-explaining context. That includes developers using Claude Code, GitHub Copilot or Codex in the ChatGPT desktop app on a codebase, and consultants and researchers with folders of PDFs and spreadsheets. It also includes small teams who want one memory per project rather than one per person per tool. It suits anyone who needs documents kept local. It is a poor fit if you need a hosted memory API to call from your own application; for that, look at Mem0 or Zep.

No. Central Brain™ does not include a chat model and does not answer questions itself. It sits beside the assistants you already use and gives them memory and file search. Claude Desktop, Claude Code, GitHub Copilot and the ChatGPT desktop app connect to it over MCP (Model Context Protocol), call its tools, and do the reasoning and writing. You keep your existing subscriptions. Central Brain™ adds the part those tools lack: a local, inspectable, shared memory and a source-traceable index of your files.

Three differences. Location: built-in memory lives in one vendor's account; Central Brain™ memory lives in files on your machine. Portability: ChatGPT memory works only inside ChatGPT and is not included in data export (source). Claude Code memory is scoped to a project through CLAUDE.md. Central Brain™ memory is read from one store by Claude Desktop, Claude Code, GitHub Copilot and the ChatGPT desktop app. Control: ChatGPT memory is capped and, since June 2026, auto-rewritten into a summary you cannot edit line by line. Central Brain™ notes are plain files you can open. Learn how these approaches differ in the AI memory layer guide.

Central memory means one memory store per project that every connected AI tool reads from and writes to, instead of separate memories inside each app. When you make a decision in a Claude session, a checkpoint goes to the project's memory files. When you open Copilot or Codex on the same project, memory_recall can return that decision. The person and their agents share the same notes. The store is on your disk, so it is also central in the sense of being under your control rather than spread across vendor accounts.

memory_checkpoint appends one established decision, fact, open question, todo or summary to the project memory. memory_recall searches the saved notes and project brief. memory_compact replaces the project brief with an updated summary supplied by the assistant and marks the included notes as compacted; it does not delete the original notes. central_brain_ping confirms that the server is running and reports the project's root folder. The connected assistant must call these tools for the workflow to work.

Central Brain™ can save five note types: decisions, facts, questions, todos and summaries. A separate project brief provides the compact baseline a new session can start from. The assistant saves individual notes with memory_checkpoint, while memory_compact updates the brief from a supplied summary. Raw chat transcripts are not automatically copied into memory, and prepared project files are indexed separately for search. People should review important notes because an assistant can save something incomplete or incorrect.

Checkpoints already written to the project memory remain available when a chat client compacts its live conversation. Compaction behavior varies by client, but it commonly summarizes or drops older conversation detail as a context limit approaches. After client compaction, or in a fresh session, memory_recall can search the saved notes and project brief. Anything that was never checkpointed may not be present. See context compaction.

Claude Desktop, Claude Code, GitHub Copilot and the ChatGPT desktop app (Chat, Work and Codex modes) connect through a project-specific MCP (Model Context Protocol) setup on Windows and Apple silicon Mac. Capabilities vary by platform and by client, because each client decides which MCP features it supports. Setup steps are in getting started. Other MCP clients may work but are not listed as supported; if you rely on one, test it before buying or ask us.

Source-file access is read-only, but the memory tools are intentionally writable. The file tools (semantic_search, search_text, read_file, list_files) search, read and list permitted project files; they do not modify, move or delete those documents. memory_checkpoint appends memory notes, and memory_compact updates the project brief and compaction state. Review the security and data-boundaries page before using sensitive material.

Yes. Central Brain™ is organized project by project. Each project has its own root folder, its own index and its own memory (five note types plus a project brief). The MCP setup is also per project, so a Claude or Copilot session attached to project A does not see project B's memory. This keeps client work, personal research and codebases apart. If you want shared context across projects, the practical approach is to keep that material in a project of its own.

Yes. The memory is stored as human-readable files on your machine, and you can open them with any text editor. This is deliberate: the readable files are the source of truth, not embeddings in a store you cannot inspect. You can check what an assistant has recorded, correct a wrong fact, or delete a note. You can also back the folder up or move it like any other project data. Hosted memory services generally offer a dashboard instead; here there is nothing between you and the notes.

semantic_search finds passages by meaning. It embeds your query with the bundled 384-dimension MiniLM model and returns the chunks whose vectors are closest by cosine similarity, so it can match text that uses different words from your query. search_text finds exact strings across files, which is better for error codes, identifiers, names and quotations. Both return results with the source path, prepared-file path, page and chunk where available. A good assistant uses semantic search for 'what do we know about X' and text search for 'where does Y appear'.

Your project content stays local by default: original files, prepared text, the local JSONL vector store, and memory notes remain on storage you choose. The bundled embedding model runs on your device, so the core preparation and search workflow does not require sending document content to a cloud embedding provider. Central Brain™ contacts Keygen for license activation and a check every 14 days, and runs air-gapped in between; Keygen receives licensing and device data, not documents. Connected AI tools receive the context they retrieve. If you configure an optional cloud connection, selected data for that connection goes to that provider. See the security page.

Central Brain™ prepares PDF, DOCX, TXT, Markdown, CSV and XLSX for text and semantic search. For images, audio and video it runs metadata workflows rather than full content extraction. OCR is optional and depends on the build; where it is included, scanned PDFs and photos can be read too. Refresh processes only files that are new, changed, removed or failed since the last run, and interrupted work can be resumed. If a format you depend on is not listed, ask support before buying.

Where the build includes it. OCR is optional and depends on the build, so check yours before relying on it. When present, OCR turns scanned PDFs and photographed pages into text that can be indexed and searched, and it runs locally like the rest of the core workflow. Accuracy depends on scan quality, so results from poor scans should be checked against the source; results keep the page number and file path where available for that purpose. If you never index scans, you can leave OCR out.

Yes. Central Brain™ runs air-gapped: preparing files, building the index, semantic and text search, reading files and the memory tools 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, for machines that never connect. Any optional cloud connection you configure (cloud embeddings, Pinecone, Postgres, Redis or a custom endpoint) needs internet, and so do cloud-hosted AI clients, which reach their models online.

Only the data needed by that connection, and only to the provider you configured. Cloud embeddings send chunk text to the embedding provider and receive vectors back. Publishing to Pinecone or Postgres sends vectors, text and metadata to that database. Redis or a custom endpoint receive whatever you point at them. Each connection is off until you set it up. Same-Wi-Fi device sharing is a separate feature: it is off by default and device trust can be revoked. Review the security page before enabling a connection for sensitive material.

Credentials for optional connections (cloud embeddings, Pinecone, Postgres, Redis, and custom endpoints) are stored locally by the installed app. The exact protection available depends on the operating system and product release, so we do not promise that every configuration uses the same encryption mechanism. The default local workflow needs no external provider keys. Sharing with other devices on the same Wi-Fi is off by default, and device trust can be revoked. See the security page.

No. Central Brain™ does not claim SOC 2, HIPAA or GDPR certification, and we will not imply otherwise. What we can state is architectural: project content is local-first by default, no cloud embedding provider is required for the core workflow, MCP file tools are read-only, projects are separated, and sharing is off by default. Licensing checks in with Keygen every 14 days (Team and Enterprise can get a build with no online check), and connected AI tools or optional cloud providers receive selected data when used. Whether that meets your compliance needs is a question for your own review. The security page lists the specifics, and you can contact us with questionnaires.

Three self-serve plans, billed monthly. Solo is $12 per month for 1 license. Pair is $20 per month for 2 licenses ($10 each). Power Pack is $35 per month for 4 licenses ($8.75 each). Team and Enterprise are custom priced; contact us about deployment support or shared infrastructure such as Pinecone or Postgres. Pricing is flat per license: no tokens, credits, retrieval quotas or reasoning tiers. Prices are as of September 2026; the current figures are on the pricing page.

Yes. Annual billing saves 15% on Solo, Pair and Power Pack. At the September 2026 monthly prices, that works out to about $122 a year for Solo instead of $144, about $204 for Pair instead of $240, and about $357 for Power Pack instead of $420. Monthly billing remains available. Check the pricing page for the exact billed amounts.

One license activates one computer running the Central Brain™ desktop app. Pair includes 2 separate one-computer keys and Power Pack includes 4, so a small team can share one plan without sharing keys. The app runs on Windows and Apple silicon Mac. There is no metering on top of the license: index as many files and make as many searches and memory checkpoints as your machine can handle, at no extra charge. If you use optional cloud services (Pinecone, cloud embeddings and so on), those providers bill you under your own accounts. Plan details are on the pricing page.

Refund terms are published on the refunds page, and that page is the authoritative version. Before buying, note that preparing files, indexing, search and memory run air-gapped, while the 14-day license check, optional cloud connections and cloud-hosted AI clients need an internet connection. If something is not working, support can often resolve it faster than a refund. Billing questions go to support@neuroaigent.com.

Team and Enterprise are custom-priced plans for organizations that need more than self-service licenses, such as deployment support or a shared index published to Pinecone or Postgres (pgvector or JSONB) in infrastructure your organization controls. Pricing is custom and depends on seat count and deployment needs. The local-first desktop model stays the same; publishing is an addition, not a replacement. To discuss requirements, use the contact page. If you only need a few seats without shared publishing, Power Pack at $35 per month for 4 licenses may be enough.

Mem0 is a hosted memory API for developers building their own agents; Central Brain™ is a desktop app for people using existing assistants. Mem0 stores AI-extracted memory summaries in its cloud, has the largest community in the category (its homepage showed 62,590 GitHub stars on September 17, 2026), SDKs and enterprise compliance. It offers a free tier and paid plans with retrieval quotas (mem0.ai/pricing). Pick Mem0 if you are shipping an application. Pick Central Brain™ if you want local files, verbatim retrieval and one memory across Claude Desktop, Claude Code, GitHub Copilot and the ChatGPT desktop app. See the AI memory layer guide for the product categories and selection criteria.

Zep is enterprise agent memory built on a temporal knowledge graph; Central Brain™ is local, file-based project memory for individuals and small teams. Zep's strengths are fast retrieval, bi-temporal facts, provenance and governance, with its Graphiti engine open source. It is metered in credits, with a free tier and paid plans (getzep.com/pricing). Pick Zep if you need graph queries over time at scale. Pick Central Brain™ if you want flat per-license pricing and notes you can open. See the AI memory layer guide for the product categories and selection criteria.

Basic Memory is the closest neighbor in philosophy: plain Markdown you own, plus full-text, semantic and knowledge-graph indexing over MCP. Its local version is free and open source, and Basic Memory Cloud is a paid hosted option (basicmemory.com). Pick Basic Memory if you want open source, a knowledge graph over your notes, or a free starting point. Pick Central Brain™ if you want a packaged desktop app that also prepares and indexes PDF, DOCX and XLSX files (and scans, where the build includes OCR), with source-traceable results and project-specific setup for Claude, GitHub Copilot and the ChatGPT desktop app. See the AI memory layer guide for the product categories and selection criteria.

Notion AI is workspace AI over your Notion pages and connected tools such as Slack and Google Drive; its memory is the workspace content itself, hosted in Notion's cloud. It is included in Notion's Business plan at $20 per member per month billed monthly, and in Enterprise, per notion.com/pricing on September 17, 2026. Notion also offers an MCP server. Pick Notion AI if your team already works in Notion. Pick Central Brain™ if your knowledge is in files on disk, you want local-first processing, or you want one memory for Claude Desktop, Claude Code, GitHub Copilot and the ChatGPT desktop app. See the AI memory layer guide for the product categories and selection criteria.