Introducing Mensara: shared memory for people and agents

I’ve spent the last three months building Mensara. It’s a shared Mind that people and their AI agents can use together, so the next person can pick up the work with the context behind it.

Watch the product explainer: Mensara, a Mind for your company.

I spent seven years as COO of Wild Earth, and I know what it costs when too much of a company’s context lives in people’s heads. Decisions, customer promises, the reasons something works the way it does. Someone always has to bring the next person up to speed.

Giving each person an AI tool doesn’t, by itself, solve that handoff. An agent can draft a good reply from the information in its conversation. The colleague who takes over still needs to know which agreement applies, what changed and where the answer came from. If that context stays in a private chat, somebody has to explain it again.

Multiplayer is at the core of Mensara. Multiple people can use the same memory, alongside the agents they already work with. People and agents are equal partners in the work: they can draw on the company’s record, contribute to it and help keep it current. You decide what each person can see and which changes need approval.

That lets the work continue across different people and tools. One person might use Claude, another Codex or ChatGPT. They can work from the same Mind, within their own access, rather than relying on someone to paste the relevant history into each conversation. The shared record holds the context that needs to survive the handoff.

A customer promise is a useful place to start. Take a simple illustration: someone in sales agrees to a one-off support exception for a customer’s launch. The person covering support wasn’t in that conversation. Before they respond, they need the agreed hours, what the exception covers, who owns it and when it ends.

The useful record contains the agreement and its source. A request for longer hours belongs there too, but as a request until someone with authority agrees to it. The next person needs to be able to tell the difference between what the customer asked for and what the company promised.

With Mensara, that person can ask their agent to read the account record and show the source behind its answer. If something is missing or contradictory, there is a specific record to check and an owner to ask. The aim is a handoff someone can act on without reconstructing the whole conversation.

The same approach applies when a teammate needs to understand why a decision was made or how a process works. Useful company memory includes the reasons and exceptions, as well as the final answer. Otherwise, a new person can follow the written instruction and still miss why it mattered.

Keeping that record current is part of the work. People and agents can suggest changes. Mensara shows the proposed wording, and your rules decide whether it can be saved straight away or needs someone to approve it. If a change is held for review, the current record stays unchanged. Once the change is applied, the next read can use it, with earlier versions retained in the history.

Sharing memory also needs boundaries. A colleague handling a support question may need the coverage agreement without needing the commercial negotiation behind it. Each person’s agent works within that person’s access. Working together doesn’t require giving everyone access to everything.

There’s a reasonable question here: couldn’t you do this with a shared folder, a wiki or the tools you already have? For some teams, those are enough. The part I want Mensara to help with is the ongoing work around the record: letting people and agents use it across tools, deciding who may change it, reviewing those changes and keeping the history.

The underlying records are plain files you can copy and keep. That matters to me. The decisions and ways of working a company builds up should remain usable beyond any one AI conversation or product. You can keep a copy of the files and their published history, readable without Mensara.

I’ve been running my own work on the system since January. Now I run Mensara on it. That gives me a place to use what I’m building every day, but the next step is to learn how it fits other teams’ real work.

We’re starting with a small founding group, and we’ll help each team get set up. I want to start with something concrete: one recurring question, one customer commitment or one piece of work that someone else needs to take over. We can then see whether the shared record helps, what still needs explaining and whether the team comes back to it.

If that sounds familiar, join the Mensara waitlist and tell us about your team.