Pieces is a background app that captures what you see, copy and, if you opt in, hear across the apps and sites you already use, saving it as a searchable memory kept on your own computer by default. It has no free plan, so what decides whether it suits you is whether you want a personal, on-device recall layer badly enough to pay from the first day of a 7-day trial.
What is Pieces?
Pieces is made by Mesh Intelligent Technologies, Inc., headquartered in Cincinnati, United States. It runs as a desktop app on Windows, Mac and Linux.
Captured memory is saved on-device by default; an optional, off-by-default encrypted cloud backup can sync it to Pieces-managed cloud storage.
What Pieces does for your day
Pieces is not an email client or calendar app; instead it captures what is on screen in apps like Gmail and Outlook as memory sources, alongside Chrome, Slack, Notion, Figma and dozens of others. With opt-in audio capture switched on, it also picks up context from meetings and conversations in Google Meet, Zoom and Microsoft Teams, including decisions, action items and call takeaways.
You can ask Pieces about any of this in plain language, scoping by time, source or topic to find or curate what you need. It has no dedicated task list; it is a memory and recall layer rather than a task manager.
Working with other people
Individual use is the default, and the Enterprise plan adds organization-wide capture policy and shared, governed memory controls so a team's context can become shared knowledge. Pieces has no public developer API; instead it offers its own MCP server, which carries your captured history into Claude, Cursor, Codex, Antigravity and any MCP-ready assistant.
Pieces pricing: what each plan gives
There is no free plan. Pro costs $18.99 per user a month, or $169.99 per user a year billed annually. Enterprise costs $22.99 per user a month, or $275.88 per user a year billed annually, and adds bring-your-own-provider API keys and organization-wide capture controls.
Every plan starts with a 7-day free trial that requires a card and can be cancelled any time before it ends; a separate 14-day money-back guarantee also applies. Pro includes switching between Claude, Gemini, ChatGPT and local models under one plan, along with unlimited MCP access, from the first day of the trial.
Privacy and security
Pieces does not train on your content, personal information or data; its own models are trained on synthetic datasets instead. It is SOC 2 Type II certified, with regular security audits and independent penetration testing, and it processes personal data on the legal bases set out in the EU and UK GDPR.
Support runs through email, a Discord community, GitHub Issues and a bookable call, alongside self-serve documentation.
Who Pieces suits
Pieces suits people who want a searchable memory of what they have seen, copied and worked on across their apps, kept on their own computer rather than a vendor's cloud, and who are ready to pay from the start rather than test on a free tier.
It suits less well anyone who wants a free plan, since the only option is a 7-day trial that requires a card, or anyone looking for a task manager or email client, since Pieces only captures and recalls; it does not manage tasks or send mail.
Frequently Asked Questions
How much does Pieces cost?
There is no free plan. Pro costs $18.99 per user a month, or $169.99 per user a year billed annually, and Enterprise costs $22.99 per user a month, or $275.88 per user a year billed annually. Every plan starts with a 7-day free trial that requires a card.
Which AI models does Pieces work with?
Pro lets you switch between Claude, Gemini, ChatGPT and local models under one plan, with unlimited MCP access included from the first day of the trial.
Does Pieces capture meetings?
With opt-in audio capture switched on, Pieces picks up context, decisions, action items and takeaways from meetings in Google Meet, Zoom and Microsoft Teams.
Does Pieces train on my data?
No. Pieces does not use your content, personal information or data to train its models; its own models are trained on synthetic datasets, and captured memory is saved on your own computer by default.
