Braid 2.1 Memory Capsules: How Local AI Context Moves Between Computers

Intersignal released Braid 2.1 on October 30 with Memory Capsules, a way to carry reviewed context between local AI conversations and computers. The release announcement targets developers running their own models who want project decisions to survive beyond one chat.

How Braid Memory Capsules work

According to the product documentation, a capsule contains a title, mission, decisions, constraints and next steps. Capsules begin excluded from chat. Users must select them and permit their inclusion; editing a capsule clears that inclusion.

A local .braidmemory export can be reviewed and imported elsewhere. It does not automatically synchronize machines or authorize an MCP client. The file is readable text, and its checksum detects changes rather than proving authorship. A request receipt shows which context was supplied, not whether the model understood it.

That last distinction is useful. A record of what the model received is easier to investigate than a promise that it remembers. If an assistant contradicts a project decision, the operator can first check whether the relevant record entered the request before blaming the model's reasoning.

Braid 2.1 price and platform limits

The local app is free; optional membership costs $99 a year. Managed WAN access is early access and requires separate activation. Models and inference servers are separate.

Apple silicon has a bundled app, but it is not notarized. Windows and Linux use Python launchers. Linux execution remains unqualified, and existing profiles are not automatically migrated.

Those limits matter more for an operational deployment than the breadth of a launch claim. Someone evaluating a local workspace should distinguish an available package from a platform combination that has actually been exercised. This is a release to inspect against a concrete workflow, not evidence of production reliability across every supported launcher.

Portable memory still needs an owner

Consider a project whose deployment constraints change. A portable note helps another machine pick up the new decision, but a second copy can still preserve the old one. Teams need a way to decide which revision is authoritative and when stale context should stop being used.

The interesting direction here is making context an inspectable artifact rather than an invisible side effect of chat history. Portability is useful when the person moving the context can also see, approve and withdraw its inclusion. It does not remove the work of keeping that context correct.

I'm Paras. I work at the intersection of AI, engineering and distribution.

I help technical founders turn product launches and engineering details into clear stories developers can understand and share.

Book a 30-minute call.