# Atlaso > The memory layer for the AI era. Atlaso gives your AI tools a shared, > persistent memory: it captures what matters as you work, recalls it before > every prompt, and carries it across Claude Code, Cursor, Codex, Claude > Desktop, OpenCode, and Antigravity (ChatGPT and Meta AI support coming). Atlaso is a hosted service with thin connectors per tool. Install with one command — no account needed first, because the command opens the browser to create it: macOS/Linux: curl -fsSL https://atlaso.ai/install.sh | bash Windows: irm https://atlaso.ai/install.ps1 | iex That installs the atlaso CLI, links the machine, then asks which of your AI tools should share the memory and wires up each one. GUI chat apps (Claude Desktop, claude.ai, any MCP client) need no terminal — they add https://mcp.atlaso.ai/mcp as a connector and authorize in the browser. You can also sign in at app.atlaso.ai/sign-in first if you prefer the dashboard. Free plan: 1 device + 1 active tool, automatic capture & recall, personal + per-project memory scopes, cloud sync, unlimited memories, web dashboard. Pro ($10/month, monthly-only, no trial): unlimited devices & tools sharing one memory, Ambient Memory (session-start orientation), nightly background enrichment, Ask-your-memory, JSON export. Build ($25/month): the developer platform — everything in Pro plus a hosted memory API for your own app or hardware: a private, isolated memory per end-user (subjects), project API keys for server-to-server use, per-project usage dashboard. Docs at https://docs.atlaso.ai/build. Secrets are scrubbed client-side before storage; memory is never trained on or sold. ## Product - [Home](https://www.atlaso.ai): what Atlaso does and how it works - [Pricing](https://www.atlaso.ai/pricing): Free plan + Pro at $10/month + Build at $25/month — full comparison - [Pricing for AI agents](https://www.atlaso.ai/pricing.md): the same plans, limits, and billing terms as plain markdown - [Install Atlaso](https://docs.atlaso.ai/install): the one-line install for macOS, Linux and Windows, and exactly what it does - [CLI reference](https://docs.atlaso.ai/cli): every atlaso command — connect, setup, status, sync, unlink, disconnect - [Docs & Help](https://docs.atlaso.ai): per-tool setup, the one-active-tool rule on Free, plans & billing behavior, troubleshooting, GUI connectors, privacy - [Docs index for LLMs](https://docs.atlaso.ai/llms.txt): machine-readable docs index (per-page .md links) - [Full docs for LLMs](https://docs.atlaso.ai/llms-full.txt): the complete documentation as one plain-text file - [What is an AI memory layer?](https://www.atlaso.ai/what-is-an-ai-memory-layer): the definitional explainer - [Atlaso vs mem0](https://www.atlaso.ai/vs/mem0): reader-matched benchmark comparison, honest limits included - [Atlaso vs ChatGPT memory](https://www.atlaso.ai/vs/chatgpt-memory): built-in memory vs a cross-tool memory layer, from official docs - [Atlaso vs Claude memory](https://www.atlaso.ai/vs/claude-memory): claude.ai memory, Claude Code memory, and what a shared layer adds - [Atlaso vs Zep](https://www.atlaso.ai/vs/getzep): Zep is agent-memory infrastructure developers build on; Atlaso is a product you install for the tools you already use - [Contact](https://www.atlaso.ai/contact): message the team (support@atlaso.ai) ## Memory for your tool - [Claude Code](https://www.atlaso.ai/for/claude-code): full automatic recall + capture on every turn - [Cursor](https://www.atlaso.ai/for/cursor): automatic capture; recall via a generated rules file - [Codex](https://www.atlaso.ai/for/codex): full automatic memory in Codex CLI and Desktop - [Claude Desktop](https://www.atlaso.ai/for/claude-desktop): memory via custom connector (MCP) — no terminal needed - [OpenCode](https://www.atlaso.ai/for/opencode): full automatic memory via a local plugin the CLI installs - [Antigravity](https://www.atlaso.ai/for/antigravity): full automatic memory in IDE and CLI - [Any MCP tool](https://docs.atlaso.ai/tools/custom-mcp): connect any MCP-capable client (Zed, Cline, …) via https://mcp.atlaso.ai/mcp ## Atlaso Build — developer platform - [Build overview](https://docs.atlaso.ai/build): give your own app, agent, or hardware a hosted memory — what Build is and how it fits - [Build quickstart](https://docs.atlaso.ai/build/quickstart): first memory stored and recalled in a few minutes - [Build API reference](https://docs.atlaso.ai/build/api-reference): the developer memory API endpoints - [Authentication](https://docs.atlaso.ai/build/authentication): project API keys and device credentials - [Subjects](https://docs.atlaso.ai/build/subjects): a private, isolated memory per end-user - [Devices](https://docs.atlaso.ai/build/devices): enrolling hardware and per-device credentials - [Limits](https://docs.atlaso.ai/build/limits): rate limits and plan quotas ## Research - [Research index](https://www.atlaso.ai/research): published research on AI memory, with full methodology and raw run logs - [Atlaso vs mem0 benchmark](https://www.atlaso.ai/research/memory-benchmark): reader-matched head-to-head on LongMemEval-S (n=500) — Atlaso +6.6pp over mem0 (p≈0.007), three independent judges, losses published - [Giving AI a subconscious](https://www.atlaso.ai/research/ai-subconscious): the Ambient Memory essay - [Adaptive stigmergy gates](https://www.atlaso.ai/research/adaptive-stigmergy-gates): when multi-agent coordination should turn on - [Where the memory lift comes from](https://www.atlaso.ai/research/growing-field-substrate): 94% retrieval, 6% format ## Legal - [Privacy Policy](https://www.atlaso.ai/privacy) - [Terms of Service](https://www.atlaso.ai/terms) - [Refund & Cancellation Policy](https://www.atlaso.ai/refunds)