Isolation before features
Multi-tenancy is close to impossible to retrofit. We built the boundary first and added capability on top of it, which is slower to start and much harder to catch up with.
About
MeshTale is made by Codas Labs. We run several products and several clients through the same set of AI tools, which is precisely how we found this problem.
Every AI tool now offers memory. Almost all of it assumes one person with one set of things to remember. That assumption holds right up until you are working across more than one client, or one team, or one company — and then it quietly stops holding, usually without telling you.
We built MeshTale because we needed it. The interesting problem was never storing the context; that part is well understood. The interesting problem is answering “which of this was that assistant allowed to see?” and being able to prove the answer afterwards.
Multi-tenancy is close to impossible to retrofit. We built the boundary first and added capability on top of it, which is slower to start and much harder to catch up with.
Every claim on this site is checked against what the product actually does, and the ones that fail get cut rather than softened. The security page has a section listing what we will not claim.
Licensed under FSL-1.1-MIT. You can read it, run it yourself, and satisfy yourself that the boundary is where we say it is.
We speak the open protocol your tools already use. Two tools, one sentence each, no software library to adopt, nothing proprietary to unpick if you leave.
Pre-launch. The memory layer works end to end against real data across a wide range of sources, and the governance surfaces are built but still switching on workspace by workspace. The pricing page marks exactly which of those are live today.